SPATIAL SCALES OF MUSKOX RESOURCE SELECTION IN LATE WINTER
A
THESIS
Presented to the Faculty
of the University of Alaska Fairbanks
in Partial Fulfillment of the Requirements
for the Degree of
MASTER OF SCIENCE
By
Kenneth J. Wilson, B.A.
Fairbanks, Alaska
May 1992 SPATIAL SCALES OF MUSKOX RESOURCE SELECTION IN LATE WINTER
By
Kenneth J. Wilson
RECOMMENDED:
Advisory Committee Chair
Department Head
APPROVED: Dean, Coll£ge-e£>Naturai Sciences
Dean orthe Graduate School
Date ABSTRACT
I examined resource selection by muskoxen in late winter on the coastal plain of the
Arctic National Wildlife Refuge, Alaska, by comparing use and availability at regional, meso, local, and micro spatial scales. Use of vegetation types for feeding appears to be based on selection of areas of shallow soft snow with high cover of sedges, dead vegetation, and total vegetation, and on selection against areas of little vegetation cover or deep hardpacked snow. Muskoxen used moist sedge, tussock sedge, and Dryas terrace tundra in proportion to availability and avoided barren ground, partially vegetated, riparian shrub, and Dryas ridge tundra. Selection for areas of shallow snow occurred within vegetation types as well as between vegetation types. Occurrence of sedges and grasses in the diet was greater than availability. Feeding zones were primarily on windblown vegetated bluffs; these areas are distributed in narrow bands along creeks, rivers, and the coastline.
i i i TABLE OF CONTENTS
ABSTRACT...... iii
TABLE OF CONTENTS...... iv
LIST OF FIGURES...... vii
LIST OF TABLES...... ix
ACKNOWLEDGEMENTS...... xi
INTRODUCTION...... 1
Effects of Snow on Arctic Ungulates...... 1
H y p o th eses...... 4
Resource Selection: A Question of Scale...... 4
STUDY A REA ...... 9
Herd History...... 9
Terrain and Vegetation Types...... 11
C lim a te ...... 13
METHODS...... 15
Foraging Site Distribution...... 15
Use and Availability of Vegetation Types ...... 18
Regional Scale Selection...... 19
Meso and Local Scale Selection...... 19
Vegetation Characteristics...... 20
Diet Selection...... 22
Environm ental V ariables...... 24
Snow Conditions...... 24
iv V
Meso and Local Scale...... 24
Microscale...... 26
Multivariate Modeling ...... 27
Selection Within Vegetation Types...... 27
RESULTS ...... 29
Foraging Site Distribution...... 29
Use and Availability of Vegetation Types ...... 29
Regional Scale Selection...... 29
Meso Scale Selection...... 33
Local Scale Selection...... 33
Test of Potential Bias due to Patch S ize...... 35
Vegetation Characteristics...... 37
Total Vegetation Cover...... 37
Cover of Forage Classes...... 37
Cover of Species...... 41
Diet Selection...... 41
Late Winter Fecal Samples...... 41
Winter-Type Samples...... 43
Rumen Samples...... 45
Environmental Variables ...... 45
Meso and Local Scale...... 45
Snow Conditions...... 45
Meso Scale...... 45
Local Scale...... 47 vi
Microscale...... 47
Multivariate Modeling ...... 50
Effect of Snow on Resource Selection...... 52
Snow and Vegetation Characteristics by Vegetation Type...... 53
Selection Within Vegetation Types...... 57
DISCUSSION...... 61
Selection Model...... 66
LITERATURE CITED ...... 72
PERSONAL COMMUNICATIONS...... 80
APPENDIX A: Cross Reference of Vegetation/Land Cover Types...... 81
APPENDIX B: Barter Island Snow Depths Over Five Years ...... 82
APPENDIX C: Vegetation Types of Used and Unused Areas Before Pooling...... 83
APPENDIX D: Cover to Biomass Regression Analyses...... 87
APPENDIX E: Rank Order of Most Abundant Species ...... 88
APPENDIX F: Diet Composition as Determined by Microhistological Analysis...... 89 LIST OF FIGURES
Figure Page
1 Schematic of spatial scales used in sampling muskox resource selection. 6
2 Map of study area and coastal plain of the Arctic National Wildlife Refuge. 10 3 Sampling design for muskox winter foraging sites in late winter of 1989 and 1990 on the coastal plain of the Arctic National Wildlife Refuge. 16
4 Map of locations of muskox foraging sites that were sampled. 30
5 The proportion of vegetation types in muskox feeding zones compared to proportions in adjacent zones, nonadjacent zones, and the study area. 31
6 Paired observations of meso and local scale vegetation type selection. 34
7 Paired observations of vegetation type selection in nested halves of muskox feeding zones. 36
8 Total vegetation cover in feeding, adjacent, and nonadjacent zones. 38
9 Cover of forage classes in feeding, adjacent, and nonadjacent zones. 39
10 Dead vegetation and abiotic cover in feeding, adjacent, and nonadjacent zones. 40
11 Percent use of forage classes as determined by composition of composite samples of muskox feces minus percent availability as determined by vegetation cover in feeding zones. 44
12 Importance (percent use x percent availability) of forage classes as determined by composition of composite samples of muskox feces and vegetation cover in feeding zones. 44
13 Snow depth and total vegetation cover of muskox feeding zones by vegetation type. 54
14 Snow depth and total vegetation cover of nonadjacent zones by vegetation type. 55
15 Snow depth and total vegetation cover of adjacent zones by vegetation type. 56
16 Snow depth and total vegetation cover of feeding, adjacent, and nonadjacent zones. 58
vii v iii
17 Conceptual model of foraging energetics based on total vegetation biomass (generalized to potential metabolizable energy intake) and snow depth (generalized to energy expenditure of cratering). 68
18 Conceptual model of selection of vegetation types by snow depth and vegetation biomass in late winter. 68
19 Hypothesized vectors of selection for vegetation types during early, mid, and late winter based on snow depth and to ti vegetation biomass. 70
20 Expected availability vectors during winters of extreme snow accumulation, after heavy use of vegetation, and the effect of a combination of both factors. 70
21 Paired observations of vegetation types of transect halves at the meso 84 scale before and after pooling.
22 Paired observations of vegetation types of transect halves at the local 85 scale before and after pooling.
23 Paired observations of vegetation types of nested halves of feeding 86 zones before and after pooling. LIST OF TABLES
T able Page
1 Definitions of spatial scales of muskox selection measured on the coastal plain of the Arctic National Wildlife Refuge in late winter of 1989 and 1990. 7
2 Chi-square analysis of regional scale selection (vegetation types of muskox feeding zones versus 492 points distributed systematically in the study area) in late winter. 32
3 Mean composition of late winter and winter-type muskox fecal samples, and mean composition of late winter rumens of hunter-killed muskoxen. 42
4 Snow depth in feeding, adjacent, and nonadjacent zones. 46
5 Snow hardness in feeding, adjacent, and nonadjacent zones. 48
6 Snow depth in muskox feeding microsites and in unused portions of paired feeding zones. 49
7 Snow hardness in muskox feeding microsites and in unused portions of paired feeding zones. 49
8 Meso scale selection (paired halves of muskox feeding zones and nonadjacent zones) for vegetation cover, snow depth, and snow hardness within the two most commonly used vegetation types. 59
9 Local scale selection (paired halves of muskox feeding zones and adjacent zones) for vegetation cover, snow depth, and snow hardness within the two most commonly used vegetation types. 57
10 Cross reference of vegetation/land cover types of four habitat studies on the coastal plain of the Arctic National Wildlife Refuge, Alaska. 81
11 Snow depth in moist sedge tundra on Barter Island during March and April during the first year of study and four previous years and the deviation from the five year mean. 82
12 Regional scale selection for vegetation types before and after pooling. 83
13 Coefficients of determination (r^) and significance levels (P) for regression analyses of the relationship between cover and biomass for major plant species on the coastal plain of the Arctic National Wildlife Refuge. 87
14 Rank order of the most abundant species by total occurrences. 88
ix X
15 Composition of late winter muskox feces collected in feeding zones as determined by microhistological analysis of plant fragments. 89
16 Composition of winter-type muskox feces as determined by microhistological analysis of plant fragments. 89
17 Percent diet composition of muskoxen killed in late winter by hunters as determined by microhistological analysis of rumen contents. 90 ACKNOWLEDGEMENTS
Funding for this research was provided by the U.S. Fish and Wildlife Service,
Alaska Fish and Wildlife Research Center through the Alaska Cooperative Fish and
Wildlife Research Unit as part of the 1002 Research Study Plans, Research Work Order
29. Funds for travel to the 1991 Arctic Ungulate Conference were provided by the Alaska
Cooperative Wildlife Research Unit, the University of Alaska Graduate School, and the
University of Alaska Vice-Chancellor for Research.
Special thanks are given to the members of my committee: Dr. David Klein, chairman, Pat Reynolds {ex efficio), and Drs. Terry Bowyer, Stephen MacLean, Thomas
McCabe, and Dana Thomas. Dave always seemed to be confident in my ability even when I wasn't. Pat was the one who introduced me to muskoxen in the wild. Terry kept me up to date. Steve taught me how to do science. Tom helped me to be realistic. Dana gave me clarity and understanding.
I would like to acknowledge the help and cooperation of many U.S. Fish and
Wildlife Service employees, especially botanists Nancy Felix and Janet Jorgenson, pilots
Dave Sowards and Roger Kaye, and dispatcher Paul Lhotka. I am especially thankful for the dedication of the volunteers who helped me during the summer field season including
Chad Boyd, Mark Biddlecomb, and especially Christian Nellemann, Mike Harris, and
Katharine Grant. Without their assistance this project would have been impossible.
Christian's enthusiasm was limitless. Mike was willing to stick it out through any conditions. Katharine had more interest than anyone else that I've met. I also appreciate the assistance of Phillip Kugzruk and Nate Collin during the winter field season. Each of them
xi x ii
contributed immensely to my knowledge and understanding of northern life, not to mention what they taught me about snowmachine repair.
I appreciate the expert flying of the pilots from Temsco Helicopters, especially Doc and his ability to fly us out of any situation. I'm thankful for the help and friendship of many of my fellow graduate students, particularly Steve Fleischman who helped with the study design and Mike Smith for his help with computers. Thanks are extended to Norma
Mosso and Sally Manning for their help with equipment and purchasing.
I am thankful for the hospitality and friendship that was extended towards me from the people of Kaktovik, particularly the Simon Tagarook family, Sam Gordon, the Daniel
Akootchook family, Susan Gordon, the James Lampe family; I thank the young people for fun games of volleyball. Hank Ramirez and other members of my Fairbanks family supported me consistently as well as my friends at Dry Creek. I wish to thank my parents,
Dave and Barbara Wilson, and my brother, Ray, for all of their encouragement. Mom always said "Go ahead, it's a once in a lifetime experience!" And she was right. The value of this experience to me goes far beyond the pages of this thesis... INTRODUCTION
Most studies of habitat selection by herbivores assume that animals are distributed according to an ideal-free distribution (Crawley 1983) where the best habitats are occupied most and the poorest habitats are least. There are many reasons, however, why this distribution may not occur. Few species are distributed solely in relation to food availability, and fewer still have a perfect knowledge of food distribution (Crawley 1983).
Also, the size, interspersion, and juxtaposition of habitat patches may result in increased distances between favorable patches limiting the ability of herbivores to exploit the best patches (Owen-Smith and Novellie 1982) by increasing the energy required for locomotion. Thus, limits on knowledge of the availability of habitats and ability to move into habitats may alter what an animal perceives as being available. Short-term foraging decisions may be based on what an animal perceives as available within some smaller radius rather than upon availability over an entire range. The purpose of this study was to assess habitat selection by muskoxen (Ovibos moschatus) in late winter on several scales that were biologically relevant and to compare selection of resources among these scales.
Additional objectives were to identify variables that affect selection by assessing snow and vegetation characteristics of used and unused areas at several different scales, and to assess diet selection.
Effects of Snow on Arctic Ungulates
Arctic ungulates inhabit ranges that are snow covered for up to eight months (Thing
1977). One of the effects of snow is to impede detection of forage (Bergerud 1974),
1 2
thereby limiting knowledge on which foraging decisions are based. Another effect of snow on arctic ungulates is to increase the energetic costs of locomotion (Fancy 1986), which results in a greater cost of moving into or between habitat patches. Snow cover also increases the cost of obtaining forage under snow, which limits availability of forage
(Fancy and White 1985, Thing 1984). These consequences of snow increase the time and energy required for foraging, decrease the ability of northern ungulates to select forage and to locate habitats where forage is most available (Adamczewski et al. 1988, Parker 1978,
Skogland 1978), and thereby reduce the average quality of forage consumed. Additionally, because of plant senescence, winter forage is typically of low quality in terms of available nutrients and digestibility (Chapin et al. 1986, Klein 1990).
Winter can influence ungulate populations in both density-dependent and density independent ways. Winter severity can limit ungulate population size independent of density such as when ice crusts and deep snow lie directly over vegetation for extended periods (Lent 1978) causing starvation. Snow and winter conditions also can influence population size through density-dependent relationships when less severe but chronic winters limit availability of food year after year (Mech et al. 1987, but see Messier 1991) or when winter ranges become depleted due to overgrazing (Klein 1968). Thus, winter snow cover and the ability of animals to cope with it are major selective forces on arctic ungulates.
Muskoxen exhibit specific behavioral, morphological, and physiological adaptations to winter (Klein 1986). They have twice the rumen capacity of caribou
(Rangifer tarandus), a slower rate of passage of digesta through the rumen, and muskoxen are able to digest forage of lower quality. They have a wide mouth that is characteristic of bulk feeders (Klein 1986). However, muskoxen also have comparably short legs and small 3
hooves that result in a reduced ability to dig in and to move through deep snow. Behavior of muskoxen in winter is energetically conservative, resulting in low rates of movement
(Jingfors 1980, Reynolds 1990a) and feeding (Jingfors 1980). Because of these adaptations, muskoxen theoretically are less constrained by forage quality, but are more constrained by geographic distribution of habitat types and snow characteristics. Muskoxen would be expected to select for concentrated areas of resources with shallow snow.
Lent and Knudson (1971), and Lent (1974) observed muskoxen digging in snow
< 30 cm deep with an integrated hardness up to 1518 kg cm on Nunivak Island, Alaska.
These animals fed on the perimeter of the island in windblown areas where snow was absent or easily excavated by fracturing it; they avoided the interior of the island where snow was deep. Likewise, Smith (1984) observed that muskoxen rarely attempted to forage through > 30 cm of snow on Nunivak Island. Rapota (1984) observed muskoxen feeding in tussock tundra on the Taimyr Peninsula, Soviet Union, until snow depth was
>20-30 cm and snow density was > 0.25g/cm. At snow depths greater than this threshold, muskoxen moved into more windswept but less productive, upland habitats with less snow cover. Thomas and Edmonds (1984) noted that muskoxen cratered in snow drifts > 50 cm deep on eastern Melville Island. M. Raillard (pers. comm.) reported that muskoxen at
Sverdrup Pass, Ellesmere Island, fed in habitats with shallow snow in March until late
April when temperatures rose to near freezing. This temperature change caused a decrease in snow hardness and muskoxen selected areas of deeper snow and more lush vegetation. 4
H ypotheses
The first objective of this study was to determine what vegetation types were selected by muskoxen for feeding. The null hypothesis tested was that use of vegetation types would be in proportion to their availability.
The second objective was to determine why particular areas were selected for feeding. The null hypotheses tested were that areas used for feeding would not differ from unused areas in: 1) snow depth; 2) snow hardness; or 3) vegetation biomass. These hypotheses were tested within vegetation types as well as among vegetation types to determine if selection for feeding areas was a function of vegetation type alone. Also, because the two years of this study differed significantly in snow depth, there was an opportunity to compare foraging patterns in deep versus shallow snow.
Because snow characteristics are determined partially by topography as it relates to wind speed and direction, muskox feeding zones were expected to be of a slope and aspect that exposed them to strong winds. These sites were expected to have greater microrelief that would provide more potential feeding microsites. And, because vegetation characteristics often are correlated with other variables such as slope, aspect, elevation, and moisture regimes, feeding zones were expected to differ from unused areas in one or more of these environmental variables.
Resource Selection: A Question of Scale
Selection for areas of shallow snow by caribou has been reported at several scales.
LaPerriere and Lent (1977) recognized three scales: 1) broad areas-selection of particular 5
valleys with shallow snow where caribou spent at least part of the winter; 2) selection of
foraging sites-sites of shallow snow within valleys where caribou concentrated their
cratering activities; and 3) microsites-selection within foraging sites for specific locations
where snow was the most shallow where caribou dug craters. At all scales selection
operated progressively toward shallower snow depths. Similarly, Lent and Knudson
(1971) described selection at two scales for muskoxen on Nunivak Island; 1) selection for
winter range on the island perimeter where snow was much shallower than in the interior of
the island; and 2) selection for foraging sites on the tops of exposed sand dunes where
snow was on average shallower than between sand dunes.
Selection at each scale, however, may not always be based on the same criteria. For instance, selection for particular vegetation characteristics may not occur on the same scale as selection for snow conditions. Underlying geographic patterns that determine the spatial distribution of the habitat characteristics of interest to muskoxen will determine selection patterns observed at each scale. In this study, availability of resources was measured at four levels and use was measured at two. Comparisons between these levels were made at four selection scales (Fig. 1). Use levels were feeding microsites and feeding zones (a zone containing the microsites). Levels of availability were feeding zones, paired adjacent zones
(an unused zone which was immediately adjacent to and surrounding feeding zones), paired nonadjacent zones (a nearby unused zone that was 100 meters beyond the adjacent zone), and the entire study area. Because selection at each scale may be based upon different criteria, predictions of the hypotheses were tested at the following spatial scales of use and availability; regional (macro), meso, local, and micro (Table 1). At the regional scale, for example, what was used by muskoxen was defined as the feeding zone and what was available was the entire study area. 6 Regional (Macro) Scale
Figure 1. Schematic of spatial scales used in sampling muskox resource selection in late winter of 1989 and 1990 in the Arctic National Wildlife Refuge. Selection at each scale was inferred from what was used for feeding versus what was available but unused. 7
Table 1. Definitions of scales of muskox selection measured on the coastal plain of the Arctic National Wildlife Refuge in late winter of 1989 and 1990. X's indicate at which scales the numbered variables were examined.
Regional Meso Scale Local Scale Microscale Scale Use Level Feeding Feeding Zone Feeding Zone Microsite Zone Availability Level Study Area Nonadjacent Adjacent Zone Feeding Zone Zone 1. Vegetation Type X XX 2. Vegetation Cover X X 3. Snow Depth X X X 4. Snow Hardness X XX 5. Environmental Variables X X 8
Because adjacent zones were contiguous with feeding zones, muskoxen were able to freely move into and have knowledge of what was available in them. If no selection occurred at this local scale due to no preference or incomplete use, adjacent zones would have a greater probability of being similar to feeding zones because of their close proximity. Because nonadjacent zones were farther away from feeding zones, muskoxen would have had less knowledge of their forage resources. If there was no selection at the meso scale, nonadjacent zones would have a lower probability of being similar to feeding zones because they are farther away. Comparisons at the local scale offer a more critical test by having a greater likelihood of showing selection if it occurred. Comparisons at the meso scale are more likely to show differences but these are less likely to be due to selection.
Because the exact criteria important to muskoxen were unknown, selection for forage was tested using a hierarchical approach that examined vegetation components at several different resolutions. STUDY AREA
The study was conducted on the coastal plain of the Arctic National Wildlife
Refuge, Alaska, in the area including and between the drainages of the Jago and Kongakut rivers (Lat. 69°30' to 70°08', Long. 142°00' to 144°00') (Fig. 2). This area was chosen because of its use by muskoxen and its accessibility by snowmachine from the abandoned
Distant Early Warning Line Station at Beaufort Lagoon, which was used as a base of winter operations.
Herd History
Muskoxen are thought to have inhabited the entire arctic coastal plain and adjacent foothills of Alaska prior to the 1900's (Spencer and Lensink 1970). The last known individuals were killed near Barrow between 1850 and 1860 (Bee and Hall 1956). In 1930,
34 muskoxen captured in Northeast Greenland were brought to Fairbanks, where they were kept until 1935-36, when 31 muskoxen were released on Nunivak Island (Spencer and Lensink 1970). The population of the introduced herd grew, and in spring 1969, 52 muskoxen from an expanding Nunivak Island population were released at Barter Island; in the following year 11 more were translocated to the Kavik River drainage west of the refuge (Burris and McKnight 1973).
After an initial period of high mortality and dispersal, a group became established in the Sadlerochit River valley (Jingfors 1980). Two other small groups dispersed to the
Canning-Tamayariak and the Jago-Okerokovik river drainages. These small groups, with a limited number of breeding-age cows, increased slowly at first, followed by a period of
9 10
Figure 2. Map of study area and coastal plain of the Arctic National Wildlife Refuge, Alaska. The study area partially overlaps with the portion of the coastal plain designated as the 1002 Area by the Alaska National Interest Lands Conservation Act. 11
rapid population growth in the late 1970's and early 1980's (Jingfors and Klein 1982). By
the mid-1980's, muskoxen began to disperse away from established areas within the Arctic
National Wildlife Refuge toward Canada to the east and beyond the Canning River to the
west By 1986, the population within the refuge reached a peak of 399 individuals
(Reynolds 19906). Since that time, the population within the refuge has remained stable at
about 350 individuals. The number of animals to the east and west of the refuge boundary,
however, has continued to increase and the herd has expanded in distribution (Reynolds
19906).
Terrain and Vegetation Types
The study area includes three major types of terrain. The hilly coastal plain consists of gently rolling terrain with slighdy elevated ridges and flat or gendy sloping areas (Everett
1982). The typical vegetation of this terrain type is moist to wet tundra. Foothills cover the southern half of the study area and are mosdy well-drained moist tundra. Floodplains are interspersed with the other two terrain types along major rivers that extend from the Brooks
Range north to the Beaufort Sea. The vegetation of this type is more complex and may include unvegetated or partially vegetated gravel bars, vegetated terraces, and vegetated bluffs. Walker et al. (1982) provide an extensive review of these vegetation types; their classification is based on moisture, dominant plant species, and plant growth forms.
Appendix A is a cross reference of vegetation types defined in this study with previous ones (Jingfors 1980, Robus 1981, O'Brien 1988, and Christiansen et al. 1990). Plant nomenclature follows Hulten (1968). 12
Wet sedge tundra occurs on poorly drained soils where standing water is common early in the growing season. The dominant plant species are mosses Sphagnum( spp.), sedges ( Eriophorum angustifolium and Carex aquatilis), and willow (Salix planifolia).
Moist sedge tundra occurs in moderately well-drained areas along the northern edge of the foothills and along drainages. The most common taxa include sedges Eriophorum( angustifolium, E. vaginatum, and Carex bigelowii), prostrate shrubs (Dryas integrifolia and Salix reticulata), forbs (Pedicularis spp. and Polygonum bistorta), and mosses
(Tomenthypnum nitens and Hylocomnium splendens).
Tussock sedge tundra occurs primarily in the foothills in well-drained upland areas.
The dominant species are the tussock-forming sedge (Eriophorum vaginatum var. spissum), mosses (Sphagnum spp.and Polytrichum spp.), prostrate shrubs (Salix planifolia, Betula nana, Vaccinium vitis-idaea, Salix phlebophylla,), Ledum and palustre the sedgeCarex bigelowii. The shrub tundra type occurs on south-facing slopes in the foothills and is dominated by erect shrubs (Salix spp., Betula nana, and Alnus crispa) and forbs such asLupinus arcticus..
The Dryas ridge type occurs on crests of ridges and in areas of frost boils that are wind swept and dry. The dominant species are Dryas integrifolia, Salix phlebophylla, and
Oxytropis nigrescens,.
Partially vegetated tundra can be interspersed with any of the other vegetation types and is dominated by nonliving cover such as rock, soil, or gravel with a variety of plant taxa. The barren ground type occurs in dry river beds and in alpine areas and is made up of gravel bars and talus slopes with little or no vegetation.
The riparian grass forb gravel bar type is made up of forbs (Astragalus alpinus,
Epilobium latifolium, Oxytropis maydelliana, Hedysarum mackenzii, Artemesia arctica, 13
Stellaria spp.) and grasses ( Festuca spp. and Agropyron spp.) in areas that were colonized recently. The riparian shrub type is composed of willows(Salix alaxensis, Salix glauca, and Salix lanata) along with grasses and forbs (Oxytropis borealis and Lupinus arcticus).
In the southern part of the study area, riparian willows grow up to 3 m in height, but closer to the coast they rarely reach 1.5 m in height
The Dryas terrace type is located on dry flat ground that is elevated and adjacent to rivers in areas with an underlying layer of gravel. The dominant species are Dryas integrifolia, with mosses (Polytrichum spp. and Dicranum spp.), small forbs {Astragalus umbellatus), prostrate shrubs (Salix reticulata, Arctostaphylos rubra), and Equisetum variegatum. Often, there are small, open canopy willows growing singly.
The coastal plain is made up of Quartemary sediments overlying sedimentary rocks that range in age from Precambrian to Quaternary (Clough et al. 1987); however, most is of poorly consolidated Tertiary and Cretaceous shale, siltstone, and sandstone bedrock. The entire study area is underlain by permafrost. Soils are of two major types, pergelic cryaquolls and pergelic cryaquepts (Rieger et al. 1979).
Climate
The climate is arctic maritime (Clough et al. 1987) and is characterized by extremely low winter temperatures with strong surface winds, short cool summers, and low annual precipitation. Records from the nearest weather station on Barter Island from 1959 to 1988
(N.O.A.A. 1989) show February to have the coldest mean temperature of the year (-28.8°
C). Average temperatures for March and April are -26.3 and -18.4°C, respectively. July is the warmest month, with a mean temperature of 4.4°C. The marine surroundings of Barter 14
Island have a greater moderating effect in the summer than in winter when the Beaufort Sea is covered by ice. At points further inland, the climate is more continental, with greater extremes in temperature.
Surface winds are consistently from either the east or west during winter
(N.O.A.A. 1988). From 1949-1985, observations of wind direction during March indicate that 32 % of the time winds were from the east, 30 % were from the west, and <10 % were from any other single direction. Winds are strongest in January (mean 24.1 km/h) and least in July (17.2 km/h). Annual precipitation averages 15.9 cm. Because mean monthly temperatures are below freezing for eight to nine months per year (Brown et al. 1975), 65 to 80 % of the precipitation falls as snow, usually through cyclonic disturbances moving eastward from the Bering Sea or from the Siberian coast (Benson 1982). Resulting snow cover on the arctic coastal plain is of the tundra-snow type, and usually consists of a hard, high density, wind-packed and sculptured layer overlying a coarse, low-density depth hoar layer (Benson 1982). Snow cover can last for up to nine months.
Data on snow depth collected by the U.S. Fish and Wildlife Service from 1985
1989 on Barter Island (Appendix B) indicate that depth in 1989 was much greater than average. These same data were not available for 1990; based on field observations it appeared that the snowfall was much closer to the long-term average during that year.
Depth measured in the same locations during the two years of this study averaged 28.0 cm
(S.D.=3.81) in 1989 and 18.7 cm (S.D.=5.70) in 1990. Mean snow hardness measured in the same locations was 15.3 kg (S.D =7.95) in 1989 and 10.8 kg (S.D.=3.00) in 1990.
Snow showed similar patterns of drift accumulation but was significantly shallower in
1990 than in 1989, but snow hardness was not significandy different. METHODS
Foraging Site Distribution
Winter field work began on 15 March 1989 and continued for four weeks; in 1990
it began on 8 March and continued for six weeks. To determine the distribution of foraging
sites, muskox groups were located within the study area from fixed-wing aircraft at the
beginning and midpoint of each field season. Flights followed a general north-south
orientation centered over major rivers and creeks and passed over known wintering areas
and locations where muskoxen had been sighted by U.S. Fish and Wildlife Service
personnel during the preceding February survey (P. Reynolds pers. comm.). Locations
were recorded on 1:63,360 scale USGS topographic maps. I assumed that major groups of
animals would occur near known wintering areas (P. Reynolds pers. comm.) or within
sight of the major drainages (Jingfors 1980, Robus 1981, Gamer and Reynolds 1986, and
O'Brien 1988). Because the study area is treeless with little topographic relief and had a complete cover of snow, muskox groups were highly visible during late winter, especially in upland areas between drainages. All radio-collared individuals in the study area as determined by previous surveys (P. Reynolds pers. comm.) were identified as being members of the groups that were located during this study.
Locations where muskox groups had been observed from the air were then relocated from the ground by the use of snowmachines. Foraging sites in the area where groups had been observed were identified and sampled unless they were obscured by accumulation of windblown snow. A foraging site (Fig. 3) was defined as an area that
15 16
Foraging Site
Figure 3. Sampling design for muskox winter foraging sites in late winter of 1989 and 1990 on the coastal plain of the Arctic National Wildlife Refuge. Foraging sites consisted of a feeding zone containing feeding microsites, surrounded by unused adjacent and nonadjacent zones. Zones were sampled using subsamples along a randomly oriented transect that extended through the entire foraging site. Each zone was therefore represented by two nested halves. 17
contained a feeding zone (a zone that muskoxen had selected for feeding as evidenced by feeding craters or feeding microsites), an unused adjacent zone, and an unused nonadjacent zone. A feeding zone was defined as an area where muskoxen fed that was 50 m or greater in radius and was separated from other feeding zones by a minimum of 200 m. It was made up of all feeding microsites within the smallest circle. A feeding microsite was defined as a continuous area of snow disturbance caused by efforts of one or more muskoxen to locate and obtain food by digging. Feeding microsites were easily distinguished from areas of snow disturbed by locomotion or social activity by determining if vegetation had been exposed and fed upon. Adjacent zones were unused areas at least 50 m wide that surrounded a feeding zone. Nonadjacent zones were unused areas that were at least 50 m wide and were 100 m from paired adjacent zones.
The foraging site was defined as the experimental unit because it contained an area where feeding had taken place that was surrounded by areas where no feeding had taken place. Although foraging sites on a regional scale were clumped, most foraging sites were widely spaced (500 m or more) and in consequence probably represent independent feeding events that took place on separate days. The average distance between foraging sites was of the same magnitude as group movement rates in late winter on the Sadlerochit River that averaged less than 0.7 km/day (Jingfors 1980). The number of individuals or groups that used foraging sites could not be precisely determined.
Use and Availability of Vegetation Types
During winter, a single, randomly oriented transect was established for each foraging site (Fig. 3). The transect passed through the center and extended to both edges of 18
each feeding zone. The length of the transect varied with the size of the area that muskoxen
had used. The adjacent zone was defined by continuing the transect in both directions for a
distance equal to one half of the feeding zone through an area that had not been used for
feeding but was contiguous with the feeding zone. After extending the transect in both
directions 100 m beyond the adjacent zone, the transect was continued through the
nonadjacent zone. The nonadjacent zone was an area that had not been used for feeding and
was defined as being the same width as half of the feeding zone. Each zone within each
foraging site was therefore represented by two nested subsamples, termed nested halves
(Fig. 3).
In July of each year of the study, coinciding with the period of peak vegetation
biomass, the transects were relocated. Because of plant phenological progression on the
arctic coastal plain from south to north (Whitten and Cameron 1980), transects were
sampled progressively from the foothills to the coast
Along each transect nested halves of each zone were classified according to
physical features and dominant vegetation type using Walker's (1983) hierarchical method.
Nested halves were categorized into units that described specific vegetation types
(corresponding to Level D of the Walker classification). Nested halves were classified by
their moisture regime, the dominant plant species in each layer of the canopy, the dominant
plant growth forms, and an overall physiognomic descriptor. Complexes of more than one
vegetation type were classified by the type that composed most of the stand. These units
were subsequently collapsed into 11 vegetation types (described in the Study Area, Terrain
and Vegetation Types section).
Regional Scale Selecdon.—Data on availability of vegetation types in the entire study area were obtained from Christiansen et al. (1990). These data consisted of a 19
systematic sample of 492 points located in the field between the Sadlerochit and Kongakut
rivers that were classified to air photo interpretable units (level C of Walker's
classification). Data were collected at 41 sites with a grid size of 12.2 km. At each site, 12
locations spaced 400 m apart were sampled. These vegetation types were collapsed into the
same 11 vegetation types reported previously, with all categories with two or fewer
observations pooled into an "other" category. Because availability of vegetation types was
estimated, a nonmapping chi-square technique (Marcum and Loftsgaarden 1980) was used
to determine if use was proportional to availability at a regional scale. Unless the grid size
from vegetation sampling was autocorrelated with some regular, underlying pattern, the
results of this test are unbiased. Both nested halves of the feeding zone were tested
separately; if results were identical for each half, then the results of only one test were
reported in a table.
Meso and Local Scale Selection.—If muskoxen fed randomly at the meso and local
scales, vegetation types chosen for feeding would be: 1) the same as the vegetation types in
paired unused areas; or 2) different from the types in paired unused areas but with no
consistent pattern of selected or avoided types. The first null hypothesis was addressed by
examining the number of times feeding zone halves (Fig. 3) were the same vegetation type
as paired adjacent and paired nonadjacent zone halves. The chi-square test of symmetry
(Agresti 1990 353:354) for paired samples was used to test the second null hypothesis
using a transition matrix that compared the vegetation types of feeding zone halves with
paired observations of vegetation types of nonadjacent zone halves. For this test it was
necessary to use a reduced matrix with any category with two or fewer observations in either a row or column pooled into an "other" category. To examine the effect of habitat 20
patch size, both null hypotheses were examined by comparing nested halves within the feeding zone.
Vegetation Characteristics
In July of each year of the study, vegetation cover was subsampled on each of the transects using a vertical point frame (Barbour et al. 1987) containing ten pins spaced 10 cm apart The point frame was centered at 5-m intervals perpendicular to each transect A cover "hit" was recorded each time a lowered pin intersected a plant or other object
Vascular plants encountered were identified to species. Unidentified specimens were collected and identified later according to Hultdn (1968). If floral parts or other distinguishing parts could not be obtained, the specimen was identified only to genus or forage class. Nonvascular plants were identified to growth form, genus, or species when possible. Nonliving cover was classified as either dead vegetation, water, soil, sand, gravel, rock, or feces. Cover of tussocks and hummocks also was recorded. Once a layer of continuous moss, lichen, dead vegetation or nonliving material was encountered, no additional hits were recorded for that pin. If dead vegetation was the only hit recorded for a pin, then it was recorded as "litter alone."
Cover was the sum of all pin hits by species for each feeding, adjacent, and nonadjacent zone divided by the total number of pins used in each zone, giving the number of hits-per-pin for each species. Usually > 200 pins were used to characterize each zone.
Cover data was examined at three different resolutions: 1) total vegetation cover, 2) cover by forage class, and 3) cover by individual species. 21
Total vegetation cover was the sum of all live plant cover for each zone. A randomized-block analysis of variance (ANOVA; Zar 1984) was used to test for a meso scale effect and a local scale effect on total vegetation cover with foraging site as the blocking variable. Each year was analyzed separately, then the years were combined. SAS
General Linear Models procedures (SAS Institute 1985) were used in all ANOVA and
MANOVA analyses. All continuous data sets were tested for normality using the
Kolmogorov-Smimov goodness-of-fit test (Zar 1984) and Q-Q plots of residuals (Johnson and Wichem 1988). A Bartlett-Box test (Neter et al. 1985) was used to test for homogeneity of variances. Rank transformations were used when the assumptions for parametric tests could not be met.
The cover of forage classes was determined for deciduous shrubs (excluding willows), evergreen shrubs (including Dryas ), forbs, grasses, horsetails, lichens, mosses, dead vegetation, willows, sedges, abiotic cover, and other (which included plants such as algae). A randomized-block multivariate analysis of variance (MANOVA) was used to test for a meso scale effect and a local scale effect on cover of forage classes with site as the blocking variable. Each year was analyzed separately and then the years were combined.
All cover variables at the species level were ranked from highest to lowest according to their total number of occurrences in all feeding, adjacent, and nonadjacent zones (this included vascular and nonvascular plants, dead vegetation, and abiotic species).
Those species that occurred in 25 or more zones ( of 129 zones: 43 feeding zones; 43 adjacent zones; and 43 nonadjacent zones) were analyzed for a meso scale effect and a local scale effect using a randomized-block ANOVA with foraging site as the blocking variable.
To control experimentwise error with a large number of variables (48), the Bonferroni inequality (Miller 1980) was used with an alpha=0.001; this yielded an experimentwise 22
error rate of alpha=0.0488. Using this approach, all statements hold simultaneously with at
least 95 % confidence. This method guards against concluding that selection occurred when
it did not. The power of the individual tests, however, is reduced as the probability of a
type I error is reduced, therefore actual P-values are reported in an appendix.
Diet Selection
Plant types that were fed upon by muskoxen were identified by forage class or
species when possible. Six to ten fresh fecal pellets were collected from each of a minimum
of ten fecal groups encountered in feeding zones during late winter. Fecal samples were combined with samples from all feeding zones within a 10 km radius that may have been
used by the same group of muskoxen because of the possibility that fecal material might not be representative of the feeding zone in which it was collected due to the slow rate of passage of digesta through the gut (mean tract retention time= 104.9 hours as reported by
Adamczewski and Chaplin et al. in press). During summer 1988, "winter-type" fecal samples (O'Brien 1988) were collected opportunistically from areas adjoining Pokok Bay,
Camden Bay, the Sadlerochit River, and the Nularvik River (Fig. 2). These samples were likely produced during winter based on their size and shape, but their age was unknown.
Fecal samples of unknown age were collected in willow stands on the Kongakut River during late winter 1990. Plant composition by forage class was estimated from microhistological analysis of feces (Todd and Hansen 1973) using 200 views per slide at the Wildlife Habitat Management Laboratory, Washington State University. Forage classes delineated were willows, shrubs (including all deciduous and evergreen shrubs other than willows), forbs, sedges, grasses, horsetails, mosses, lichens, and other. 23
To determine availability of forage, cover of forage classes in feeding zones was
averaged across feeding zones used each year that were within a 10 km radius of each
other. Abiotic cover and dead vegetation (because it was not identified to species) were
excluded and the number of hits/pin was rescaled to 100 % to allow for comparison to use
of forage. Selection for forage classes was determined by testing the null hypothesis that
use of forage classes was not different from availability by using a randomized-block
MANOVA with muskox group as the blocking variable. Also, importance values (Bowyer
and Bleich 1984) were calculated for each forage class by multiplying use by availability
and rescaling to 100 %. The Quade test (Conover 1980:295, Alldredge and Ratti 1986) was
used to rank order forage classes based on the difference between use and availability. The
same test was used to rank order forage classes based on importance values. The null
hypothesis tested was that selection for all forage classes was equal.
Rumen samples were provided by the Alaska Department of Fish and Game and the
U.S. Fish and Wildlife Service from eight hunter-killed adult muskoxen (one cow and
seven bulls). These samples were obtained during the March-April hunting seasons, 1982
1989. One sample from an adult cow that died in poor physical condition in 1982 was also
obtained. These samples were analyzed for plant fragment composition (Wildlife Habitat
Management Laboratory).
Environmental Variables
Slope, aspect, wetness, and microrelief were determined at subsampling points at
5-m intervals along the transects. Slope categories were coded as: 1 = flat; 2 = moderate; and 3 = steep. If slope was "moderate" or "steep", aspect was recorded as 1 of the 8 points 24
of the compass. Wetness categories were: 1 = dry, 2 = moist, 3 = wet, and 4 = very wet.
Terrain microrelief (degree of brokenness within one meter) was categorized on a scale from 1 to 5 with 1 = smooth to 5 = extremely broken.
Mean values of slope, microrelief, and wetness were calculated for each feeding, adjacent, and nonadjacent zone. Data on aspect were converted into values defined according to x and y rectangular coordinates where the x coordinate (sine) describes east (x
= 1) to west (x = -1) aspects and the y coordinate (cosine) describes north (y = 1) to south
(y = -1) aspects. Mean sines (Xn) and cosines (Yn) were calculated according to Zar
(1984:428). The mean aspect of each of the feeding, adjacent, and nonadjacent zones was determined as being the angle having a sine = Xn and cosine = Yn. Environmental variables were tested for differences at the meso and local scales using sign tests (Conover
1980:122).
Snow Conditions
Meso and Local Scale.—Snow depth and snow hardness (Lent and Knudson 1971) were subsampled at points on 5-m intervals along the transect (Fig. 3) in feeding, adjacent and nonadjacent zones during late winter. Snow depth was measured to the nearest centimeter. The Rammsonde penetrometer (Lent and Knudson 1971), which measures vertical resistance of snow layers to penetration, was used to measure hardness to the nearest kilogram at each subsampling point When a subsampling point fell within a feeding crater or other area of disturbed snow cover, another point in the nearest 25
undisturbed snow was measured. Because the transect extended in two directions, each
zone was represented in the sample by two nested halves. Transects were marked with
stakes so that they could be relocated the following summer.
Snow accumulations > 100 cm could not be measured accurately and were
conservatively rounded to 101 cm. Data on snow collected over what was later determined
to be a permanent body of water were deleted from the data set because of the lack of
potential forage at such sites. Values of snow hardness (R) were calculated using the
equation: R = Whn/X + W + Q (Lent and Knutson 1971), where: R = snow hardness value
(kg); W = weight of drop hammer (kg); h = height of hammer drop (cm); n = number of
hammer blows; X = penetration after n blows (cm); Q = weight of penetrometer (kg).
Hardness was analyzed as snow hardness (kg); however, it can be integrated over the entire snow profile to obtain integrated snow hardness (in kg cm).
A randomized-block ANOVA was used to test for meso and local scale effects on snow depth with foraging site as the blocking variable. The same analysis was used to test for meso and local scale effects on snow hardness. Each year was analyzed separately, then the years were combined.
A null model, if selection for shallow snow or soft snow had not occurred, predicts that the frequency of times that feeding zones are shallower or softer than paired adjacent or paired nonadjacent zones would be equal to 0.5. To test this hypothesis, a chi-square goodness-of-fit test was used with Yates' correction for continuity (Zar 1984:48) with 1 degree of freedom.
Unpublished data from snow surveys conducted on Barter Island from 1985 to
1989 were obtained from the U.S. Fish and Wildlife Service to examine year to year 26
variation in snow depth. Also, ten of the 1989 feeding zones were resampled in 1990 to examine differences between years of the study. A repeated measures ANOVA was used to examine the effect of year on depth and hardness in the resampled feeding zones.
Microscale.—Subsamples of snow depth were taken at feeding microsites within each foraging site. The depth of least disturbed edges of craters (LaPerriere and Lent 1977) was measured. If a crater was large, presumably because more than one animal had used it, several subsamples were collected. If the surface of the ground was uneven where muskoxen had fed in areas of hummocks, tussocks, mounds, or ridges of low center polygons, snow depth was measured to the tops of these features because most forage was located there. If the snow cover had been completely removed by feeding, then similar, undisturbed features nearby were measured for snow depth. Mean microsite depth was calculated for each foraging site.
A method to measure snow hardness at microsites was used in 1990 under specific conditions when selection by muskoxen for microsites could be simulated. These were met primarily when muskoxen fed in tussock sedge tundra. In those cases, hardness was measured among undisturbed tussocks that were visible from the surface. Mean microsite hardness was calculated for each foraging site.
A randomized-block ANOVA was used to examine the effect of microsite scale
(microsite vs feeding zone) on both depth and hardness with foraging site as the blocking variable. For snow depth, years were examined separately and then combined. 27
Multivariate Modeling
A correlation matrix of snow variables, environmental variables, cover of selected plant species (species known to have been used as forage), and cover of abiotic species was calculated at each scale. When variables had a significant Pearson correlation coefficient greater than 10.701 (Neter et al. 1985), one of the variables was eliminated from further analysis. The remaining variables were analyzed with a stepwise discriminant analysis using the SAS STEPDISC procedure (SAS Institute 1985). Variables that best discriminated between feeding zones and unused zones at the meso and local scales were identified and the jackknifmg procedure (SAS DISCRIM procedure) was used to estimate the percent of samples correctly classified.
Selection Within Vegetation Types
To determine if selection for feeding zones was a function of selection for vegetation types or for characteristics that varied within vegetation types, paired observations of used (feeding zone halves) and unused areas (either adjacent or nonadjacent zone halves) of the same vegetation type were examined for differences in snow depth, snow hardness, and total vegetation cover at the meso and local scales. These matched pairs were further partitioned by vegetation type into three groups: tussock sedge tundra, moist sedge tundra, and other. If two matched pairs occurred on the same transect (each scale considered separately), one was randomly chosen to be included in the analysis and the other was eliminated to avoid pseudoreplication. A randomized-block MANOVA was 28
used to test for a meso scale effect and a local scale effect in tussock sedge tundra and moist sedge tundra with foraging site as the blocking variable. RESULTS
Foraging Site Distribution
In 1989, one solitary bull and six groups (128 individuals) were observed in the study area from the air. Twenty foraging sites from four or more groups were later located on the ground (Fig 4). In 1990, six groups (117 individuals) were observed in the study area from the air, and a total of 24 foraging sites used by five or more groups were later located on the ground.
Use and Availability of Vegetation Types
The vegetation types most commonly used by muskoxen for feeding (Fig. 5) were moist sedge tundra (37 %) and tussock sedge tundra (37 %). Muskoxen also fed in Dryas terrace (9 %), and riparian grass forb gravel bars (7 %). They used wet sedge (5 %), partially vegetated tundra (2 %) and shrub tundra (2 %) to a lesser extent and were not observed feeding in riparian shrub, Dryas ridge, barren ground, or water cover types.
Regional Scale Selection.—Thirty-eight% of the study area was composed of moist sedge tundra, and 22 % was tussock sedge tundra. Observed use of vegetation types differed significantly from expected use (P=0.012) (Table 2, Appendix C). Ninety-five percent simultaneous confidence intervals showed that use of moist sedge, tussock sedge, and Dryas terrace types did not differ from availability. The "other" category was used less than expected. This result was due to avoidance of riparian shrub, water, Dryas ridge, barren ground and very low use of wet sedge, shrub, and partially vegetated tundra.
29 30
Figure 4. Map of the locations of muskox foraging sites that were sampled in March and April of 1989 and 1990 in the Arctic National Wildlife Refuge Figure 5. The proportion of vegetation types in muskox feeding zones compared to proportions in adjacent zones, nonadjacent zones, and in the study area in the Arctic National Wildlife Refuge, Alaska in March and April, 1989 and 1990. 32
Table 2. Chi-square analysis of regional scale selection (use of muskox feeding zones versus expected use based on availability of vegetation types in the study area) on the coastal plain of the Arctic National Wildlife Refuge in late winter of 1989 and 1990. A nonmapping chi-square technique (Marcum and Loftsgaarden 1980) was used for a reduced matrix where any category with two or fewer observations in either a row or a column were pooled in to an "other" category. See Appendix C for data before pooling and analysis of other nested half.
Feeding Zones Type Observed Expected
Other 7a 15.0 Moist Sedge 16 16.2 Tussock Tundra 16 9.8 Dryas Terrace 4 2.0
d.f.=3, Chi-square= 11.1, P=0.012 a = Use < Availability (P for simultaneous confidence intervals < 0.05). 33
Meso Scale Selection.—If muskoxen fed randomly, vegetation types chosen for feeding would be: 1) the same as the vegetation types available nearby; or 2) different from the types available nearby but with no consistent pattern of selected types or avoided types.
The first null hypothesis was addressed by examining the percentage of similarity: nested halves of the nonadjacent zone were the same vegetation type as nested halves of the feeding zone 44.2 % and 42.5 % of the time (67.4 and 42.5 % after pooling). The second null hypothesis was addressed using a transition matrix (Fig. 6 and Appendix C) that compared the vegetation types of feeding zone halves with paired observations of vegetation types of nonadjacent zone halves. Illustrated graphically, the null model predicted symmetry about the diagonal in Fig. 6. An example of symmetry is if muskoxen fed in tussock sedge tundra and avoided riparian shrub at one foraging site, and at another foraging site they fed in riparian shrub and avoided tussock sedge. No symmetry between use and availability was found (chi square test of symmetry P=0.03), therefore use was significantly different from what was expected based on the null model. Muskoxen were more likely to feed in moist sedge tundra, tussock sedge tundra or Dryas terrace, and to avoid wet sedge, shrub tundra, Dryas ridge, or partially vegetated tundra.
Local Scale Selection.—If muskoxen fed randomly, vegetation types chosen for feeding would be: 1) the same as the vegetation types available adjacent to the feeding zone; or 2) different from the types available adjacent to the feeding zone but with no consistent pattern of selected types or avoided types. Because of their close proximity to adjacent zones, feeding zones had a high probability of being the same vegetation type if selection did not occur. Therefore, comparing similarities at this scale was a more critical test of selection. 34
Figure 6. Paired observations of meso scale vegetation type selection (types used for feeding versus types available in nonadjacent zone) and local scale vegetation type selection (types used for feeding versus types available in adjacent zone) by muskoxen in late winter of 1989 and 1990 in the Arctic National Wildlife Refuge. The chi-square test for symmetry (Agresti 1990) was calculated for a reduced matrix where any category with two or fewer observations in either a row or a column were pooled into an "other" category. Observations in shaded cells indicate no difference between zones. See Appendix C for data before pooling and analysis of other nested half.
Meso Scale
Nonadjacent Zone Feeding Zone A B C D A Other i 1 2 0 B Moist Sedge 14 ...4 C Tussock Tundra 4 1 . 9 D Dryas Terrace 1 ...... d.f.=6, Chi-square=13.9 P=0.03
Local Scale
Adjacent Zone Feeding Zone A B C D A Other 1 1 B Moist Sedge 7 C Tussock Tundra 7 1 m d D Dryas Terrace 3 d.f.=6, Chi-square=12.5 P=0.05 35
Percent similarities of vegetation types at the local scale were slightly greater than at
the meso scale. Halves of the feeding zones were of the same type as the adjacent zones
48.9 and 52.4 % of the time (55.8 and 69.0 % after pooling). A test of the second null hypothesis using the transition matrix approach (Fig. 6, Appendix C) revealed that use was significantly different from expected at this scale (chi-square test of symmetry P=0.05).
Feeding zones were more often in moist sedge tundra, tussock sedge tundra, and Dryas terrace and less often in the other types, whereas adjacent zones were more often in water, wet sedge, shrub tundra, riparian shrub, Dryas ridge, and barren ground. Note that riparian shrub was available adjacent to feeding zones but was not used for feeding (Fig. 5).
Test of Potential Bias Due to Patch Size.—Adjacent zones may have been of different vegetation types than paired feeding zones due to patch size of vegetation rather than because muskoxen chose not to feed in an unfavorable vegetation type. Because the nested halves of the feeding zone had exactly the same proximity to each other as a half of the feeding zone had to a half of the paired adjacent zone (Fig. 3), there was an opportunity to compare the similarity of vegetation types at the local scale with the similarity of vegetation types within the feeding zone. Similarity of nested halves of feeding zones were expected to be as low as similarity at the local scale if patch size was the only factor.
Nested halves of feeding zones were of the same type 95.4 % of the time (100 % after pooling). Therefore the similarity of nested halves of feeding zones was much greater than the similarity of feeding and adjacent zones. The transition matrix was symmetrical about the diagonal (Fig. 7, Appendix C), as indicated by the chi-square test of symmetry which was was not significant (P=0.92). This indicates that differences between areas selected and areas not selected at the local scale were not biased by patch size but were due to selection for favorable vegetation types and avoidance of unfavorable ones. 36
Figure 7. Paired observations of vegetation type selection in nested halves of feeding zones used by muskoxen in the Arctic National Wildlife Refuge in late winter of 1989 and 1990. The chi-square test for symmetry (Agresti 1990) was calculated for a reduced matrix where any category with two or fewer observations in either a row or a column were pooled into an "other" category. Observations in shaded cells indicate no difference between halves. See Appendix C for data before pooling.
Feeding Zone Half #2 Feeding Zone Half #1 A B C D A. Other II B. Moist Sedge M C. Tussock Tundra D. Dryas Terrace 4] d.f.=6 Chi-square=2.0 P=0.919 37
Vegetation Characteristics
To determine why selection of vegetation types in feeding zones occurred, the
characteristics that distinguished areas used and avoided by muskoxen were identified.
Vegetation characteristics were examined at three different resolutions using a hierarchical
approach to determine if there was selection for total vegetation cover, forage classes, and
individual plant species.
Total Vegetation Cover—Linear regressions that model plant biomass based on
cover for several of the most abundant species within the study area, were developed by
Felix et al. (1989) and are presented in Appendix D. Vegetative cover is assumed to index
vegetative biomass and forage availability. Total vegetation cover (Fig. 8) was greater
(randomized-block ANOVA: £=0.004) in the feeding zone (mean=1.156 hits/pin,
S.E.=0.074) than in the adjacent zone (mean=0.986 hits/pin, S.E.=0.075). This pattern
was consistent both years although stronger in 1990. Total vegetation cover in the feeding
zone (Fig. 8) was even greater (£<0.0001) compared with the nonadjacent zone
(mean=0.899 hits/pin, S.E.=0.069).
Cover of Forage Classes.—At the meso scale, cover of evergreen shrubs, sedges
(Fig. 9), and dead vegetation (Fig. 10) was greater in the feeding zone (£=0.0053,
£=0.0469, £=0.0013, respectively). Abiotic cover (Fig. 10) was significantly greater in
the nonadjacent zone (£=0.0104).
At the local scale, cover of evergreen shrubs, sedges (Fig. 9), and dead vegetation
(Fig. 10) was greater in the feeding zone (randomized-block MANOVA: £=0.0152,
£=0.0092, £<0.0001, respectively). Abiotic cover (Fig. 10) was significantly greater in the adjacent zone (£=0.002). 38
Figure 8. Total vegetation cover (the sum of the cover (hits/pin) of all vascular and nonvascular plants) in muskox feeding zones, adjacent zones, and nonadjacent zones during late winter of 1989 (n=19) and 1990 (n=24) in the Arctic National Wildlife Refuge. A minimum of 200 pins were used to characterize each sample. Bar is standard error. 1 = feeding zone significantly different (P<0.05) from adjacent zone, m = feeding zone significantly different (P<0.05) from nonadjacent zone. (L) = feeding zone significantly different (P<0.05) from adjacent zone for both years combined. (M) = feeding zone significantly different (P<0.05) from nonadjacent zone for both years combined. 39
Figure 9. Mean cover and S.E. of forage classes in 1989 (n=19) and 1990 (n=24) in muskox feeding, adjacent, and nonadjacent zones in the Arctic National Wildlife Refuge. A minimum of 200 pins were used to characterize each sample. 1= feeding zone significantly different (p<0.05) from adjacent zone, m = feeding zone significantly different (p<0.05) from nonadjacent zone. (L) = feeding zone significantly different (p<0.05) from adjacent zone for both years combined (MANOVA p=0.0011). (M) = feeding zone significantly different (p<0.05) from nonadjacent zone for both years combined (MANOVA p=0.0477). 40
Dead Vegetation Abiotic Cover Cover Figure 10. Mean cover and S.E. of dead vegetation and abiotic cover in muskox feeding zones, adjacent zones and nonadjacent zones in 1989 (n=19) and 1990 (n=24) in late winter in the Arctic National Wildlife Refuge. A minimum of 200 pins were used to characterize each sample. 1 = feeding zone significantly different (p<0.05) from adjacent zone, m = feeding zone significantly different (p<0.05) from nonadjacent zone. (L) = feeding zone significantly different (p<0.05) from adjacent zone for both years combined (MANOVA p=0.001). (M) = feeding zone significantly different (p<0.05) from nonadjacent zone for both years combined (MANOVA p=0.0477). 41
Cover of Species.—There was evidence in feeding microsites that muskoxen fed on
the following species:Salix alaxensis, S. lanata, S. plamfolia, S. phlebophylla; sedges
Eriophorum vaginatum, E. angustifolium, Carex aquatilis, C.grasses, bigelowii, forbs
Stellaria spp., shrub Dryas integrifolia when mixed in with other forage species, and moss
Hylocomnium splendens when growing with E. angustifolium. If muskoxen selected
feeding zones based on the availability of these species, then their cover would be greater in
the feeding zones. Conversely, muskoxen might have also selected feeding zones by
avoiding less preferred species or areas of greater soil, gravel, or rock cover. Unpreferred
species were expected to be species such as Betula nana, Cassiope tetragona, Ledum palustre, and Aulacomnium turgidum, which were not fed on even when they were without
snow cover in foraging sites.
These and 33 other common species were tested for differences at meso and local
scales. No significant differences occurred at the meso scale (Appendix E). At the local
scale (Appendix E), the cover of one species, the mossAulocomium turgidum, was less in
the feeding zone (randomized-block ANOVA P=0.0006). This result was expected if
muskoxen selected feeding zones based on avoidance of unpreferred species. At a significance level of P=0.05 (which increased the power to detect differences), only two
vascular plant species differed at the local scale and one at the meso scale.
Diet Selection
Late Winter Fecal Samples.—Microhistological analyses of fecal pellets collected in late winter (Table 3, Appendix F) indicate a high use of sedges (39.1 %) and mosses
(24.6 %). Use was significantly different from availability as determined by Table 3. Mean composition of late winter and winter-type muskox fecal samples collected in the Arctic National Wildlife Refuge, and composition of late winter rumens of hunter-killed muskoxen. See Appendix F for raw data.
Percent Diet Composition Sedges Grasses Willows Shrubs Horsetails Forbs M osses Lichens Other Late Winter Feces (n=9) Mean (S.E.) 39.1 (5.1) 13.9 (5.6) 9.2 (1.6) 4.1 (0.5) 0.3 (0.2) 6.7 (1.3) 24.6 (4.2) 1.6 (0.4) 0.4 (0.2)
Winter-type Feces (n=5) Mean (S.E.) 23.5 (2.4) 7.9 (1.5) 1 8 .2 (4 .2 ) 3.0(1.1) 1.9 (1.3) 8.0 (2.8) 31.1(6.1) 5.0 (1.6) 0.0 (0.0)
Late Winter Rumens (n=9) Mean (S.E.) 31.0(5.3) 18.8(3.5) 8.2 (2.3) 4.0 (1.6) 6.5 (6.5) 12.9(3.3) 14.9(1.9) 3.1 (1.4) 0.5 (0.3)
4^ N> 43
vegetative cover in feeding zones where fecal samples were collected (MANOVA
P=0.0195); sedges (P=0.0002) and grasses (£=0.0014) were selected for and horsetails, lichens, willows (P=0.0422), and shrubs other than willow (£=0.0003) were avoided
(Fig. 11).
Based on the difference between use and availability, the Quade test (Fig. 11) ranked the selection for forage classes in the following order (from greatest preference to greatest avoidance): sedges, grasses, mosses, forbs, horsetails, lichens, willows, and shrubs other than willow. The Quade test based on importance (Fig. 12) ranked forage classes in a different order: sedges, mosses, willows, shrubs other than willow, forbs, grasses, lichens, and horsetails. Sedges make up a large proportion of both the diet and the cover in feeding zones, and yet they were still selected. The different ranking of grasses by the two indices illustrates an important distinction: though selection for grasses is high, they do not make up a large proportion of the diet or the habitat. Although there seems to be neither preference or avoidance of mosses at this scale, they make up a relatively large proportion of both the habitat and the diet Lichens and horsetails do not contribute much to the diet at such low availabilities.
Winter-Tvpe Samples.—In comparison to the late winter samples, there are much lower proportions of sedges in the winter-type samples (Table 3, Appendix F). The mean proportion of willow in the winter-type samples (18.2 %) is much higher than the mean of the late winter samples (9.2 %), with the Kongakut River sample being highest. Because the exact year and season that winter-type pellets were produced is unknown, examining their composition is a much less precise way of determining forage selection during a particular part of winter. Winter-type samples may represent an average of forage selection over the entire winter. If so, their higher willow content may indicate greater use 44
Figure 11. Percent use minus percent availability for each forage class as determined by composition of muskox feces collected in feeding zones and vegetation cover of feeding zones in March and April of 1989 and 1990 in the Arctic National Wildlife Refuge. Samples from feeding zones within 10 km of each other were combined, resulting in a sample size of 8. Bars are standard errors. Selection for forage classes underscored by the same line was not significantly different (p<0.05) as determined by the Quade test. * indicates use was significantly different from availability (p<0.05 MANOVA).
Figure 12. Importance (percent use x percent availability) for each forage class as determined by composition of muskox feces collected in feeding zones and vegetation cover of feeding zones in March and April of 1989 and 1990 in the Arctic National Wildlife Refuge. Samples from feeding zones within 10 km of each other were combined, resulting in a sample size of 8. Error bars are standard errors. Importance of forage classes underscored by the same line was not significantly different (p>0.05) as determined by the Quade test. 45
of willows in early winter. However, because these samples may represent selection over a
wider time period as well as a wider geographic area, it is not possible to determine the exact cause of these differences.
Rumen Samples.—Microhistological analyses of rumen samples (Table 3,
Appendix F) indicate that sedges (mean=31.0 %), grasses (mean=18.8 %), mosses
(mean=14.9 %), and forbs (mean=12.9 %) made up most of the diets. The mean
proportion of willows used was 8.2 %.
Environmental Variables
Forty feeding zones were on or within 100 m of some type of topographic relief
and were subject to wind scaring. Twenty-eight of these were on bluffs within 100 m of a
creek or river, 5 were on bluffs along the edge of a bay, 5 were centered on small vegetated
pingos, and 2 were on exposed sides of hills. Thirty-four of the 40 feeding zones were
within 200 m of a low lying area that served as a snow catch. All 44 appeared to have been
exposed to strong winds.
Meso and Local Scale.—No pattern occurred in the differences between feeding
zones and either nonadjacent or adjacent zones in wetness, slope, microrelief, or aspect (P
> 0.1) based on sign tests at each scale.
Snow Conditions
Meso Scale.—Snow depth was shallower in feeding zones (Table 4) than in
nonadjacent zones (randomized-block ANOVA P=0.001). Feeding zones ranged from the 46
Table 4. Snow depth (cm) in muskox feeding, adjacent, and nonadjacent zones in late winter in the Arctic National Wildlife Refuge. A randomized block ANOVA was used to test for meso and local scale effects and an ANOVA was used to test for a year effect in the feeding zone.
Nonadjacent Zone Meso Scale Year Mean S.E. Range Effect Feeding Zone 1989 52 3.2 24.7-79.5 P<0.0001 Year N Mean S.E. Range 1990 34.7 2.3 19.0-79.5 P=0.001 1989 20 32.7 1.98 16.33-51.2 Overall 42.6 2.3 P=0.001 1990 24 21.2 0.91 12.2-27.0 . Overall 44 26.4 1.34 Adjacent Zone Local Scale Year Effect P<0.0001 Year Mean S.E. Range Effect 1989 47.8 2.6 16.8-65.7 P<0.0001 1990 37 2.6 11.3-71.4 P<0.0001 Overall 41.9 2.0 P<0.0001 47
shallowest depth of 12.2 in 1990 in partially vegetated tundra to the deepest of 51.2 in moist sedge tundra in 1989. Selection for feeding zones with soft snow was also evident
(Table 5). Snow hardness in the feeding zone (mean=15.5 kg) was less (randomized-block
ANOVA P=0.034) than snow hardness in nonadjacent zones (mean=18.1 kg). These patterns were consistent in both years of the study.
Local Scale.—Snow depth (Table 4) was shallower in feeding zones than in adjacent zones (randomized-block ANOVA P<0.0001). Snow hardness of feeding zones
(Table 5) was less (randomized-block ANOVA P=0.0019) than the snow hardness of adjacent zones (mean=18.9 kg). In 1989 alone, however, this difference was not significant (randomized-block ANOVA P=0.157) because of higher variability.
Microscale.—Muskoxen were observed using three different types of feeding microsites. When the surface of the ground was relatively flat and without microrelief, they dug craters. When the surface had vegetated hummocks, polygon ridges, pingos, mounds, or tussocks, they exposed vegetation by pushing snow off of the tops of these features.
And when vegetation was protruding through the snow cover, cratering was not necessary.
When muskoxen fed in microsites where vegetation was below the snow cover, snow depth of microsites (Table 6) was shallower (randomized-block ANOVA P<0.0001) than in unused portions of the feeding zones in which the microsites occurred. Mean depth of feeding microsites ranged from less than 1 cm on the tops of exposed tussocks in 1990 to 46.4 cm in craters dug in moist sedge habitats in 1989. In six of the feeding zones, willows protruding through the snow cover were heavily browsed.
When muskoxen fed in microsites where vegetation was below the snow cover, the snow hardness of microsites (Table 7) was less (randomized-block ANOVA P=0.0051) 48
Table 5. Snow hardness (kg) in muskox feeding zones, adjacent zones, and nonadjacent zones in late winter in the Arctic National Wildlife Refuge. A randomized block ANOVA was used to test for meso and local scale effects and an ANOVA was used to test for a year effect in the feeding zone.
Nonadjacent Zone Meso Scale Year Mean S.E. Range Effect Feeding Zone 1989 21.3 1.8 6.6-36.1 P=0.0296 Year N Mean S.E. Range 1990 15.5 1.2 5.8-27.2 P=0.0566 1989 20 17.9 2.12 4.3-33.5 Overall 18.1 1.1 P=0.034 1990 24 13.5 0.98 6.7-27.1 Overall 44 15.5 1.14 Adjacent Zone Local Scale Year Effect P=0.1822 Year Mean S.E. Range Effect 1989 22 2.2 6.4-43.8 P=0.157 1990 16.3 1.3 6.3-30.4 P=0.023 Overall 18.9 1.3 P=0.0019 49
Table 6. Snow depth (cm) in muskox feeding microsites (craters) and in unused portions of paired feeding zones in the Arctic National Wildlife Refuge in late winter of 1989 and 1990.
Microsite Feeding Zone Microscale Year N Mean S.E. Range Mean S.E. Range Effect
1989 13 29.7 2.6 16.5 -46.4 34.5 2.66 22.11 - 51.15 P=0.0007 1990 16 9.8 1.5 0.1 -24.9 23.0 0.61 19.5-27.0 P<0.0001 Overall 29 18.3 2.4 28.1 1.6 P<0.0001
Year Effect P=0.0001
Table 7. Mean snow hardness (kg) in muskox feeding microsites and in unused portions of paired feeding zones on the coastal plain of the Arctic National Wildlife Refuge in late winter of 1990.
Microsite Feeding Zone Microscale Year N Mean S.E. Range Mean S.E. Range Effect
1990 11 11.2 0.4 9 - 12.5 13.9 0.85 6.7 - 27.1 P=0.0051 50
than the mean snow hardness of unused portions of the feeding zones in which the microsites occurred.
Incidental observations of the effect of snow on locomotion indicated that muskoxen preferred to walk in areas where the snow was extremely shallow or where it was hard enough to support their body weight When their tracks were followed, it was apparent that they often followed wind blown ridges or meandered along the tops of hard packed snow dunes rather than walking in areas of deep soft snow. When the snow was very deep and soft, muskoxen in groups walked single file or in two rows.
Selection for shallow snow occurred at all three scales that were examined during both years of the study. Only 1 of 44 nonadjacent zones had shallower snow than paired feeding zones (chi-square with Yates' correction for continuity P<0.001) and only 3 of 44 adjacent zones were shallower than paired feeding zones (chi-square with Yates' correction for continuity PcO.OOl). There were no times when mean snow depth of microsites was greater than what was available in paired feeding zones. Selection for soft snow also occurred at all three scales as well, though not as often. Thirteen of 44 paired nonadjacent zones were softer (chi-square with Yates' correction for continuity P=0.0085) and 11 of 44 adjacent areas were softer (chi-square with Yates' correction for continuity P=0.0014).
There were no times when mean snow hardness of microsites was greater than what was available in the feeding zone.
Multivariate Modeling
Snow variables, environmental variables, cover of selected plant species (those thought to be important as forage), cover of forage classes exclusive of selected species, 51
and cover of abiotic species were examined for colinearity at the local and meso scales. The only variables with a Pearson correlation coefficient greater than 10.701 was the cover ofE.
vaginatum, which was positively correlated with microrelief and deciduous shrub cover at the local scale. The environmental variables, as sampled, were not correlated to snow depth; neither were any of them other than microrelief correlated with any of the cover variables that were included.
E. vaginatum was removed from the data set to reduce multicollinearity and then stepwise discriminant function analysis (DFA) was used to identify variables that best discriminated between feeding zones and either adjacent zones or nonadjacent zones. At the meso scale, snow depth, dead vegetation and gravel cover had significant (P<0.05) F- values and were entered into the model. At the local scale, snow depth, and cover of dead vegetation, gravel cover, and S. lanata had significant (P<0.05) F-values and were entered into the model. In a jackknifing procedure, the meso scale model correctly classified 78.6
% of the feeding zones and 80.9 % of the nonadjacent zones. The local scale model correctly classified 85.7 % of the feeding zones and 79.6 % of the adjacent zones.
These models showed that muskoxen were avoiding areas of gravel and S. lanata cover with deep snow and were feeding in areas of shallow snow with high dead
vegetation cover. Individual species that were observed to be used as forage did not enter
the models, other than S. lanata which had higher cover in the adjacent zones indicating
avoidance. Although muskoxen were observed feeding on S. lanata, this was usually when it was growing in moderate or shallow snow depths and when its stems were protruding through the snow. It appears that muskoxen were unable or unwilling to make use of this forage resource even when it was growing adjacent to where they were feeding because it generally occurred in areas of deeper snow. 52
Effect of Snow on Resource Selection
The difference in snow depth between years (repeated measures ANOVA
£=0.0009, n=10) provided an opportunity to examine foraging decisions that were made
under different conditions. As would be expected based on the prevailing snow conditions, the 1989 microsite mean (Table 6) was greater (ANOVA £<0.0001) than the 1990 microsite mean. The difference between the microsite snow depth and the feeding zone snow depth, however, was greater in 1990 (-13.2 cm) than in 1989 (-4.8 cm), indicating that microscale selection may be more stronger under shallow snow conditions.
Feeding zones (Table 4) were significantly shallower in 1990 than in 1989
(£<0.0001). The difference between the feeding zone snow depth and the adjacent zone snow depth (local scale difference), however, was nearly identical in both years (1989=-
15.1 cm, 1990 =-15.8 cm), showing that selection at this scale was not affected by the overall snowfall of the two years. The meso scale difference (as indicated by the difference between feeding zone and nonadjacent zone snow depth) was consistent in both years despite the greater snow depths of 1989.
The snow hardness of feeding zones (Table 5), though less in 1990, was not significandy different between years (£=0.18) because of greater variability in 1989. This was the same pattern in the snow hardness data from the 1989 sites that were resampled in
1990. Under deep snow conditions, hardness was more variable probably due to greater formation of sastrugi (wind sculptured snow dunes).
Total vegetation cover (Fig. 8) was significandy greater in 1990 than in 1989 in the feeding zone (ANOVA £=0.0396). The local scale difference and the meso scale difference 53
was much greater in 1990 than in 1989. It appears that when muskoxen are faced with
greater snow depths, they are not as selective for areas of greater total vegetation cover.
To determine if preferences for particular forage classes were expressed more
clearly without the constraint of snow, the cover of each of the forage classes in the feeding
zones was compared between years. There was a greater amount of forb cover in feeding
zones in 1990 (ANOVA P=0.013) than in 1989, and a greater cover of sedges (P=0.0025)
and abiotic cover (P<0.0001) in 1989 feeding zones (Figs. 9 and 10).
Snow and Vegetation Characteristics by Vegetation Type
In feeding zones, the vegetation types with the shallowest snow were partially vegetated tundra (12.2 cm, n=l) (Fig. 13) and Dryas terraces (mean=14.9 cm,
S.D.=3.51, n=4). The deepest snow occurred in shrub tundra (42.0 cm, n=l) and moist sedge tundra (mean=31.9 cm, S.D.=9.19, n=18). The vegetation type with the greatest total cover of vegetation was riparian forb grass gravel bars (mean=2.01 hits/pin,
S.D.=0.40, n=3). The vegetation type with the least vegetation cover was moist sedge tundra (mean=0.79 hits/pin, S.D.=0.20, n=18).
In nonadjacent zones the vegetation type with the shallowest snow was partially vegetated tundra (mean=30.7 cm, S.D.=15.6, n=4) (Fig. 14), and the type with the deepest snow was riparian grass forb gravel bar (85.6 cm, n=l). The lowest vegetation cover occurred in barren ground (mean=0.137 hits/pin, S.D.=0.070, n=2), and the greatest cover of vegetation was in shrub tundra (mean=1.21 hits/pin, S.D.=0.044, n=2).
In adjacent zones the vegetation type with the shallowest snow was theDryas ridge type (mean=10.8 cm, S.D.=11.1, n=3) (Fig. 15), and the type with the deepest snow was 54
Snow Depth (cm)
O Moist Sedge Tundra (n=18) • Dryas Terrace (n=4)
O Shrub Tundra (n=l) T Partially Vegetated Tundra (n= 1)
□ Tussock Tundra (n= 16) ■ Riparian Grass Forb Gravel Bar (n=3)
Figure 13. Mean snow depth and total vegetation cover of muskox feeding zones by vegetation type on the coastal plain of the Arctic National Wildlife Refuge in 1989 and 1990. Bars are standard deviations. 55
Snow Depth (cm)
O Moist Sedge Tundra (n=18) • Dry as Ridge (n=2)
O Shrub Tundra (n = 2) T Partially Vegetated Tundra (n=4)
□ Tussock Sedge Tundra (n=8) ■ Riparian Grass Forb Gravel Bar (n= 1)
A Wet Sedge Tundra (n=4) E Barren Ground (n=2)
► Riparian Shrub (n=l)
Figure 14. Mean snow depth and total vegetation cover of nonadjacent zones (unused areas near but not adjacent to muskox feeding zones) by vegetation type on the coastal plain of the Arctic National Wildlife Refuge in late winter of 1989 and 1990. Bars are standard deviations. National Wildlife Refuge in late winter of 1989 and 1990. Bars are standard deviations. standard are Bars 1990. Arctic the and of 1989 of plain winter coastal late the areas in on type (unused Refuge zones vegetation Wildlife by adjacent zones) of National cover feeding vegetation muskox to total and adjacent depth snow Mean 15. Figure Total Vegetation Cover (hits/pin) O Moist Sedge Tundra (n=12) Tundra Sedge Moist O O Shrub Tundra (n=l) Tundra Shrub O □ Tussock Tundra (n=9) Tundra Tussock □ A W et Sedge Tundra (n=5) Tundra Sedge et W A RiparianShrub (n=4) ► Snow Depth (cm) Depth Snow ▼ Partially Vegetated Tundra (n= 1)(n= Tundra Vegetated Partially ▼ E Barren Ground (n=3) Ground Barren E (n=3) Dry Terrace as • • Dry as Ridge (n=2) Dry Ridge as • iainGasFr rvlBr( l) = (n Bar Gravel Forb Grass Riparian ■ 56 57
partially vegetated tundra (84.8 cm, n=l). The lowest vegetation cover occurred in barren ground (mean=0.154 hits/pin, S.D.=0.184, n=3), and the greatest cover of vegetation was in the riparian shrub type (mean=1.30 hits/pin, S.D.=0.521, n=4).
In bivariate space (snow depth and total vegetation cover) muskoxen selected for a subset of what was available to them (Fig. 16). The smaller ellipse contains 95% of the feeding zones in bivariate space, whereas the larger ellipses contain 95 % of the adjacent and nonadjacent zones.
Selection Within Vegetation Types
To determine if selection for soft shallow snow and areas of high vegetation cover was a function of selection for specific vegetation types or selection for specific factors within vegetation types, matched pairs within moist sedge tundra and tussock sedge tundra were examined.
At the meso scale (Table 8), snow depth of matched pairs was significantly shallower in feeding zones in tussock sedge tundra (P=0.0014) and moist sedge tundra,
(P=0.0001). There was no difference in snow hardness in either of the vegetation types
(P=0.219, P=0.7383 respectively). Total vegetation cover was not significantly different in either vegetation type (P=0.7175, P=0.7022 respectively).
At the local scale (Table 9), snow depth was shallower in feeding zones for both habitat types (tussock sedge tundra P=0.0066, moist sedge tundra P=0.0009). Hardness was not significantly different in the feeding zones in either of the habitat types at the local scale (P=0.1587, P= 0.732, respectively) and total vegetation cover was not significantly 58
Snow Depth (cm)
Figure 16. Snow depth and total vegetation cover of muskox feeding zones, adjacent zones, and nonadjacent zones in late winter of 1989 and 1990 on the coastal plain of the Arctic National Wildlife Refuge. Each ellipse contains 95 percent of the observations of each zone. Muskoxen selected for feeding zones with shallower snow and greater vegetation cover compared to what was available. 59
Table 8. Meso scale selection (paired halves of muskox feeding zones and nonadjacent zones) within the two most commonly used vegetation types for vegetation cover, snow depth, and snow hardness in late winter of 1989 and 1990 in the Arctic National Wildlife Refuge.
Tussock Tundra (n=10) Feeding Zone Nonadjacent Zone Meso Scale Variable Mean S.E. Mean S.E. Effect* Snow Depth (cm) 25.6 1.591 40.062 4.5 P=0.0001 Snow Hardness (kg) 10.762 1.512 17.69 2.94 P=0.219 Total Veg. Cover (hits/pin) 1.2656 0.067 1.3036 0.07 P=0.7175 * randomized block MANOVA P=0.020
Moist Sedge Tundra (n=l 1) Feeding Zone Nonadjacent Zone Meso Scale Variable Mean S.E. Mean S.E. Effect £ Snow Depth (cm) 31.8 2.712 43.7 3.4 P=0.0014 Snow Hardness (kg) 22.285 2.307 23 1.3 P=0.7383 Total Veg. Cover (hits/pin) 0.9138 0.095 0.868 0.07 P=0.7022 £ randomized block MANOVA P=0.0017
Table 9. Local scale selection (paired halves of muskox feeding zones and adjacent zones) within the two most commonly used vegetation types for vegetation cover, snow depth, and snow hardness in late winter of 1989 and 1990 in the Arctic National Wildlife Refuge.
Tussock Tundra (n= 14) Feeding Zone Adjacent Zone Local Scale Variable Mean S.E. Mean S.E. Effect* Snow Depth (cm) 26.057 1.554 35.150 .236 P=0.0009 Snow Hardness (kg) 11.200 1.372 13.880 1.700 P=0.1587 Total Veg. Cover (hits/pin) 1.265 .071 1.310 .079 P=0.2511 * randomized block MANOVA P=0.020
Moist Sedge Tundra (n=15) Feeding Zone Adjacent Zone Local Scale Variable Mean S.E. Mean S.E. Effect £ Snow Depth (cm) 31.300 2.150 37.900 1.500 P=0.0066 Snow Hardness (kg) 21.000 1.622 23.700 2.104 P=0.732 Total Veg. Cover (hits/pin) .813 .061 .891 .098 P=0.1692 £ randomized block MANOVA P=0.018 60
different (P=0.2511, P=0.1692, respectively). Within vegetation types at both the meso and local scales muskoxen chose feeding zones based on snow depth alone. DISCUSSION
Selection for vegetation type was consistent at all three scales tested. Muskoxen
used moist sedge tundra, tussock sedge tundra, and Dryas terrace in proportion to
availability at all scales. Tussock sedge tundra, and moist sedge tundra were frequently
used. Muskoxen selected against the following cover types at all scales: water, wet sedge,
Dryas ridge, riparian shrub, barren ground, and shrub tundra. Results of this study clarify
the observations of Jingfors (1980) and Robus (1981) who reported frequent use of
riparian habitats (terrace, gravel bar, willow bar, and creek willow thicket) and dry tundra
habitats (dry ridge and tussock meadow) and infrequent use of wet sedge and wet-moist
sedge complex (heath polygon) vegetation types by muskoxen in late winter (See Appendix
A for cross reference of vegetation types). Although I observed frequent use of riparian
habitats, I did not observe muskoxen using riparian shrub communities (which includes closed- canopy willow bars and creek willow thickets) and the willows that grow there
despite their availability at the local scale. This illustrates an important distinction that was
not evident in previous studies (Jingfors 1980 Robus 1981, and O'Brien 1988) due to the
spatial scales examined. During the two years of this study, snow was too deep for
muskoxen to make use of the riparian shrub communities. Browsing on willows in areas of shallow snow, however, was evident in areas where they protruded through the snow surface, especially in Dryas terraces. Undoubtedly, the most important distinction between
Dryas ridge which was avoided and Dryas terrace which was used was the presence and use of small, open-canopy willows in the terraces, which were exposed above the snow cover due to wind action.
Selection of vegetation types for feeding appears to be based on both snow and vegetation characteristics. At both the local and meso scales, total vegetation cover (which 61 62
was assumed to index plant biomass and forage availability) was greater in the feeding
zones, suggesting that muskoxen selected for areas of greater plant biomass. Cover of
evergreen shrubs (including Dryas), sedges, and dead vegetation was greater in feeding
zones at the local and meso scale. Abiotic cover (rock, gravel, soil, water) was less in the
feeding zones at both scales.
Muskoxen appear to be selecting feeding zones at the resolution of total vegetation cover and cover of forage classes rather than cover of individual forage species. The two
approaches used in this study (multiple randomized-block ANOVA's on cover of the most
abundant species and discriminant function analysis on cover of known forage species) failed to show selection based on forage species at these scales. I hypothesize that individual species (as well as particular plant parts) are selected on a smaller scale than was measured in this study. Closer examination of feeding microsites and microhistological analysis of ingested plant species may show selection for particular species.
Once a feeding zone was selected, muskoxen appeared to further select (for ingestion) particular forage classes within feeding zones. Although muskoxen used feeding zones with significantly greater cover of sedges and evergreen shrubs, shrubs were actually ingested significantly less than their availability within the feeding zone. Shrubs may be indicators of good feeding areas, but not because they are used as forage. The shrub forage class was primarily made up of Dryas integrifolia which tends to grow in greater abundance on convex slopes. Ingestion of grasses was much greater than availability, but they made up a small portion of the diet and the habitat in contrast to mosses, which made up a large percentage of both the diet and the habitat and were ingested in proportion to their availability. Willows have been considered of primary importance to muskoxen in winter in the southern extremes of their range (Tener 1965, Jingfors 1980, Robus 1981, Klein 1986, 63
O'Brien 1988); however, willows made up less than 10 % of the late winter diets of muskoxen in the study area. It appears that muskoxen will select for willows when they are growing in areas of shallow snow and when they are protruding through the snow surface in small clumps, but will avoid them when they are in deep snow or when they are prostrate and covered with snow.
Topography, as it affects snow distribution, appears to have a large effect on selection of feeding zones although the sampling design employed in this study failed to reveal any significant difference in slope or aspect at the scales sampled. Slope and aspect alone did not adequately describe the topographic features that seemed to be most important. In most cases, feeding zones were located where the slope was convex where windblown snow could not be trapped (in contrast to concave slopes that tended to act as snow traps). This observation is supported by a study of caribou winter habitat by
Fleischman (1990) conducted in Interior Alaska, who found that late winter snow depth was positively related to vegetation height and negatively related to slope and exposure (to midday sun) and convexity of slope. Because of little topographic relief in the study area, muskox feeding zones were concentrated on vegetated bluffs along creeks, rivers, and the coastline. Bluffs along these features tend to be contiguous, less isolated and therefore more accessible. Vegetated bluffs not associated with river corridors tend to be more discontinuous and isolated from other potential feeding zones. Muskoxen that have dispersed to the west of the refuge also make use of riparian corridors (M. Biddlecomb pers. comm.).
Although no differences in microrelief were detected at the local and meso scales, it is my opinion that the presence of tussocks and vegetated hummocks and pingos allows 64
opportunities for muskoxen to select for favorable feeding microsites that have less snow cover. Whenever these features were present in feeding zones they were heavily fed upon.
Selection for areas of shallow snow is a major factor in determining the distribution of feeding zones. Selection for shallow snow operates on at least three scales toward progressively shallower depths. Although there was no difference in depth between adjacent and nonadjacent zones, feeding zones were an average of 15.5 cm shallower than adjacent zones and 16.2 cm shallower than nonadjacent zones. Microsites were an average of 9.8 cm shallower than feeding zones. Nevertheless, muskoxen in the study area contended with adverse snow conditions, especially in 1989. The greatest mean depth of craters in a feeding zone (46.4 cm, Table 5) establishes a new snow-depth threshold for muskoxen. The mean crater depth of 1989 (29.7 cm) was comparable to Smith's (1984) threshold of 30 cm of snow on Nunivak Island, and is in the upper range of Rapota's
(1984) 20-30 cm of depth which caused a shift in habitat use toward areas of shallower snow. The mean depth of feeding zones (Table 6) in 1989 was greater than both of these thresholds.
Selection for areas of soft snow also occurred at three scales. On average, feeding zones were 3.4 kg softer than adjacent zones and 2.6 kg softer than nonadjacent zones. In
1989, hardness was significantly different only at the meso scale because of higher variability, but in 1990, when snow was shallower and softer than in 1989 and muskoxen may have been less constrained in other ways, they selected for softer snow at all scales.
Discriminant function analysis models at the meso and local scales suggest that snow depth was the single variable most influential in discriminating between used and unused areas because it consistently entered into the models as the first variable selected in the stepwise proceedure. The only vegetation variable that loaded positively into DFA 65
models was dead vegetation cover. Dead vegetation cover may be a better indicator of late
winter forage availability than total vegetation cover because it represents what was
available after plant senescence. These results further support the conclusion that muskoxen
at these scales select for areas of shallow snow and high forage availability.
When snow was shallow and less constraining in 1990, muskoxen selected for
areas with greater cover of forbs and total vegetation and less abiotic cover. These differences suggest that muskoxen were better able to discriminate through shallow snow and to select between potential feeding zones based more on vegetation characteristics during that year. Selection of feeding microsites also appeared to be less constrained in
1990 because selection at the microscale during that year reduced the amount digging necessary to uncover forage much more than did microscale selection in 1989.
Based on the results of this study, the availability of potential feeding zones could be accurately predicted 50 % of the time based on vegetation type alone because some vegetation types were avoided by muskoxen because of their low forage availability and some vegetation types were avoided because of deep snow. Several studies have shown a relationship between vegetation type and snow characteristics in the Alaskan Arctic. Evans et al. (1989) noted snow to be shallow in dry habitat types and deep in wet habitat types.
Brooks and Collins (1984) described a linear relationship between depth of snow and height of vegetation. However, snow characteristics within vegetation types in late winter appear to be too variable to accurately evaluate the availability of potential muskox feeding habitats based only on vegetation type. Discrimination within a vegetation type based on snow depth also occurred. Therefore, descriptions of habitats based on vegetation type alone does not reflect the way that habitats are used by muskoxen. An example, is the foraging decision made by a group of five bulls that fed in tussock sedge tundra on a 66
northwest-facing bluff above the Ekaluakat River. In summer, both sides of the bluff appeared to be identical in vegetation composition. But in winter, the northeast side of the bluff was covered with hardpacked snow over a meter in depth while the northwest side had snow so shallow that sedge leaves from the tops of tussocks were protruding through it.
The effect of predation on muskox habitat selection in late winter in the study area is probably minimal for several reasons. Although wolves are abundant in drainages in the
Brooks Range, they have rarely been observed on the coastal plain during late winter (D.
Young, pers. comm.). Brown bears generally do not emerge and begin to prey on large prey species at this time of the year. Polar bears are present and may potentially prey on muskoxen, but they probably rarely do. And because none of the vegetation types available to muskoxen offer cover of sufficient height or density, it is unlikely that vegetation type selection is based on the availability of hiding cover.
Selection Model
Muskoxen are well adapted to dealing with energetic constraints in late winter.
They have a low metabolic rate in winter (Tyler and Blix 1990), and they have an intrinsic pattern of seasonal weight loss (Adamczewski and Gunn et al. in press), lower forage intake, and longer rumen retention time (Adamczewski and Chaplin et al. in press).
Nevertheless, there is likely to be a threshold that muskoxen face where metabolizable energy intake is less than energy expended while foraging. In other ungulates, studies have shown that not only is energy required to dig through snow to obtain forage (Fancy 1986), but the effects of foraging time lost while digging craters (which limits the daily rate of 67
forage intake) may be even more energetically severe (Fleischman 1988, Goodson et al.
1991). These factors, combined with the effect of low forage quality, set an energetic
threshold beyond which continued foraging would result in changes in body composition
severe enough to limit survival or reproduction. Such a threshold, though its exact bounds
or effects have not been defined, would likely be the basis from which muskoxen make
foraging decisions.
This study has shown that muskoxen select for areas of shallow snow and high
forage availability (Fig. 16). In theory, when forage availability is generalized to potential
metabolizable energy intake (MEI) and snow depth is generalized to potential energy
expenditure while cratering, muskoxen should maximize energy intake while minimizing
energy expenditure (Fig. 17). Although muskoxen are at a negative energy balance
throughout the winter (White et al. 1981), relative energy gain should be experienced if the
MEI is greater than the energy expended while obtaining forage underlying the snow cover.
The observed use of vegetation types in late winter is hypothesized to be a function
of selection for feeding zones farthest from this threshold (closest to the upper left side of
Fig. 17). To illustrate how selection is affected by this threshold, the characteristics of
vegetation types in the study area were generalized in Fig. 18 based on mean values of
snow depth and vegetation cover by vegetation type (Figures 13-15). Muskoxen avoided
areas that had snow and vegetation characteristics beyond the energy threshold, therefore
some vegetation types were avoided entirely, and a subset of other vegetation types within the energy threshold were used.
To optimize metabolizable energy intake in early winter, muskoxen should feed in areas of the highest vegetation biomass such as in closed riparian shrub communities. As winter progresses, low lying areas such as these will probably accumulate snow faster than 6 8
Figure 17. Generalization of total vegetation cover and snow depth of used and unused zones to potential metabolizable energy intake (MEI) and energy expenditure in the selection of feeding zones. Based on data collected in this study (Fig. 16), muskoxen are expected to select feeding zones to maximize their energy intake while minimizing their energy expenditure.
Figure is. ueneranzea moaei or vegetation types assorted oy snow aepm ana vegetation biomass in late winter based on data collected during this study (Figures 14, 15, and 16) and a probable relationship between vegetation characteristics and snow depth (Brooks and Collins 1984, Evans et al. 1989). Heavy line indicates the location of the maintenence threshold beyond which muskoxen would be expected to experience relative energy loss while foraging. 69
other vegetation types. During mid-winter muskoxen can probably select from a variety of vegetation types without a substantial gain or loss in energy balance (MEI minus energy expenditure while cratering). During a severe winter or by late winter, muskoxen are forced to select for those remaining areas where energy required for cratering is lowest (in areas with the shallowest snow), which in most cases will be areas with low plant biomass such as partially vegetated tundra. My observations are supported by Adamczewski et al. (1988) who reported that in late winter, only areas with topographic relief and shallow snow presented opportunities for caribou to forage. Figure 19 shows the probable progression in the relationship between plant biomass and snow depth in each of the vegetation types from early to late winter and selection vectors illustrate the likely optimizing strategy of muskoxen during each season or during different years of varying snow conditions. When snow depth is severe, muskoxen will “cut their losses” and move to higher, more windblown terrain with lower availability of forage.
If enough body fat is not accumulated by muskoxen during late summer and early winter, then the effect of a severe winter or overuse of winter range on survival and reproduction potential are greatly increased. The effects of deep snows, heavy use of winter range, or a combination of both will force muskoxen to forage in areas where energy cost of foraging is greater than the potential MEI (Fig. 20). At low muskox densities, a severe winter of deep snow accumulation would likely have the greatest effect on calf survival because of their limited fat reserves. At high densities, deep snows result in greater adult male mortality in winter (Gunn et al. 1989), presumably because of increased energy expenditures by bulls while competing for females during rut.
In the Arctic National Wildlife Refuge, muskoxen concentrate their feeding activities in river corridors (O'Brien 1988), as well as along streams and the coastline. 70
Figure 19. Hypothesized vectors of selection by muskoxen (arrows) during early, mid, and late winter for vegetation types based on snow depth and total vegetation. Direction of arrow indicates which vegetation types would be preferred during each winter period if muskoxen maximize energy intake while minimizing expenditure. Vegetation types are numbered according to Fig. 18.
Figure 20. Expected availability vectors during severe winters of extreme snow accumulation, during winters with heavy use of vegetation, and the effect of a combination of both factors. Tne heavy line indicates the energetic threshold in winter. 71
Winter feeding areas selected in these corridors occur on windblown bluffs where snow is shallow and vegetation is accessible. Corridors are important not only because of the quality of the habitats available, but also because of greater accessibility of those habitats because contiguous windblown bluffs allow for movement between feeding areas. Relative to the density of muskoxen in the refuge on the coastal plain, densities within corridors are high (O'Brien 1988). It is not known whether density related social or environmental pressures are responsible for the dispersal of muskoxen from traditional use areas, but the availability of habitats in late winter, and the concentration of those habitats in relatively small corridors (which may increase social pressures), may be a factor in this dispersal. LITERATURE CITED
Adamczewski, J. Z., C. C. Gates, B. M. Soutar, and R. J. Hudson. 1988. Limiting
effects of snow on seasonal habitat use and diets of caribou(Rangifer tarandus
groenlandicus) on Coats Island, Northwest Territories, Canada. Can. J. Zool.
66:1986-1996.
, A. Gunn, B. Laarveld, and P. F. Flood. (In Press). Seasonal changes in weight,
condition and nutrition of free-ranging and captive muskox females. Rangifer.
, R. Chaplin, J. Schaefer, and P. F. Flood. (In Press). Intake, digestibility and
passage rate of a supplemented hay diet in captive muskoxen. Rangifer.
Agresti, A. 1990. Categorical data analysis. J. Wiley & Sons. New York. 558pp.
Alldredge J. R. and J. T. Ratti. 1986. Comparison of some statistical techniques for
analysis of resource selection. J. Wildl. Manage. 50:157-165.
Barbour, M. F., J. H. Burk, and W. D. Pitts. 1987. Terrestrial plant ecology.
Benjamin/Cummings Pub. Co. Inc., Menlo Park, Calif. 604pp.
Bee, J. W. and E. R. Hall. 1956. Mammals of northern Alaska: on the arctic slope. Misc.
Pub. No. 8. Univ. Kansas Museum Nat. Hist. 309pp.
Benson, C. S. 1982. Reassessment of winter precipitation on Alaska's arctic slope and
measurements on the flux of wind blown snow. Geophysical Institute Rep. No. UAG
R-288. Univ. Alaska Fairbanks.
Bergerud, A. T. 1974. Relative abundance of food in winter for Newfoundland caribou.
Oikos 25: 379-387.
Bowyer, R. T and V. C. Bleich. 1984. Effects of cattle grazing on selected habitats of
southern mule deer. Calif. Fish and Game 70:240-247.
72 73
Brooks, J., Ill, and W. Collins. 1984. Snow cover and interpretation of vegetation/habitat
inventories. Pages 203-210 in V. J. Labau and C. L. Kerr, eds. Inventorying forest
and other vegetation of the high latitude and high alpine regions. Proc. Int. Symp.
Soc. Amer. For. Reg. Tech. Conf.
Brown, J., R. K. Haugen and S. Parrish. 1975. Selected climate and soil thermal
characteristics of the Prudhoe Bay Region. Pages 3-11 in J. Brown ed., Ecological
investigations of the tundra biome in the Prudhoe Bay Region, Alaska. Biol. Pap.
Univ. Alaska Spec. Rep. Ser. No. 2, Fairbanks. 215pp.
Burris, O. E. and D. E. McKnight. 1973. Game transplants in Alaska. Alaska Dept. Fish
and Game, Tech. Bull. No. 4:12-17.
Chapin, F. S., Ill, G. R. Shaver, and R. A. Kedrowski. 1986. Environmental controls
over carbon, nitrogen and phosphorus fractions inEriophorum vaginatum in Alaskan
tussock tundra. J. of Ecology. 74:167-195.
Christiansen, J. S., D. C. Douglas, and M. K. Raynolds. 1990. Comparison and
implimentation of classified vegetation maps derived from LANDSAT-TM and SPOT
satellite imagery data bases for delineating wildlife habitat availability and distribution.
Pages 80-85 in T. R. McCabe, ed. Terrestrial Research: 1002 Area - Arctic National
Wildlife Refuge, Annual Progress Rep., 1989. U.S. Fish Wildl. Serv., Anchorage,
Alaska. 168pp.
Clough, N, P. Patton, and A. Christiansen, eds. 1987. Arctic National Wildlife Refuge,
Alaska, coastal plain resource assessment - report and recommendation to the
Congress of the United States and final legislative environmental impact statement:
Washington, D.C. U.S. Fish and Wildlife Service, U.S. Geologic Survey, and
Bureau of Land Management. Vol. 1. 208pp. 74
Conover, W. J. 1980. Practical nonparametric statistics, Second ed. J. Wiley and Sons,
Inc., New York. 493pp.
Crawley, M. J. 1983. Herbivory: The dynamics of animal plant interactions. Univ.
California Press. Berkeley. 437pp.
Evans, B. M„ D. A. Walker, C. S. Benson, E. A. Nordstrand, and G. W. Petersen.
1989. Spatial interrelationships between terrain, snow distribution and vegetation
patterns at an arctic foothills site in Alaska. Holarctic Ecology 12:270-278.
Everett, K. R. 1982. Soils and terrain associations of the foothills and coastal plain portion
of the Arctic National Wildlife Refuge. Pages B-l to B-24in Proposed oil and gas
exploration within the coastal plain of the Arctic National Wildlife Refuge, Alaska.
U.S. Fish and Wildlife Service. Region 7, Anchorage, Alaska.
Fancy, S. G. 1986. Daily energy budgets of caribou: a simulation approach. PhD. Thesis,
Univ. Alaska Fairbanks.
, and R. G. White. 1985. Energy expenditures by caribou while cratering in snow.
J. Wildl. Manage. 49:987-993.
Felix, N. A., S. C. Bishop, M. K. Raynolds, S. J. Fleischman, and L. M. Koestner.
1989. Snow melt, plant phenology, and seasonal availability of forage nutrients and
biomass in concentrated and peripheral calving areas of caribou on the arctic coastal
plain, in T. R. McCabe, ed. Terrestrial Research: 1002 Area - Arctic National Wildlife
Refuge, Annual Progress Rep., 1988.
Fleischman, S. J. 1988. A model of the energy required by caribou to dig a feeding crater
(abstract), p. 186 In: Cameron, R.D. and J.L. Davis eds. Proc. Third North Am.
Caribou Workshop. Alaska Dept. Fish & Game. Juneau. Wildl. Tech. Bull. No. 8. 75
. 1990. Lichen availability on the range of an expanding caribou (Rangifer tarandus)
population in Alaska. M.S. Thesis, Univ. Alaska Fairbanks.
Gamer, G. W. and P. E. Reynolds. 1986. Arctic National Wildlife Refuge coastal plain
resource assessment: final report baseline study of the Fish, wildlife, and their habitats.
U.S. Dept, of Interior.
Goodson, N.J., D.R. Stevens, and J.A. Bailey. 1991. Effects of snow on foraging
ecology and nutrition of bighorn sheep. J. Wildl. Manage. 55:214-222.
Gunn, A., F. L. Miller, and B. McLean. 1989. Evidence for and possible causes of
increased mortality of bull muskoxen during severe winters. Can. J. Zool. 67:1106
1111.
Hultdn, E. 1968. Flora of Alaska and neighboring territories. Stanford Univ. Press.
1008pp.
Jingfors, K. T. 1980. Habitat relationships and acdvity patterns of a reintroduced muskox
population. M. S. Thesis, Univ. Alaska Fairbanks. 116pp.
and D. R. Klein. 1982. Productivity in recently established muskox populations in
Alaska. J. Wildl. Manage. 46:1092-1096.
Johnson, R. A. and D. W. Wichem. 1988. Applied multivariate statistical analysis. Second
ed. Prentice-Hall, Inc. Englewood Cliffs, N.J. 607pp.
Klein, D. R. 1968. The introduction, increase, and crash of reindeer on St. Matthew
Island. J. Wildl. Manage. 32:350-367.
. 1986. Latitudinal variation in foraging strategies. Pages 237-246 in O. Gududsson,
ed. Grazing research at northern latitudes. Plenum Publishing, New York.
. 1990. Variation in quality of caribou and reindeer forage plants associated with
season, plant part, and phenology. Rangifer, Special Issue No. 3:123-130. 76
LaPerriere, A. J., and P. C. Lent. 1977. Caribou feeding sites in relation to snow
characteristics in northeastern Alaska. Arctic 30:101-108.
Lent, P. C. 1974. Final report: ecological and behavioral study of the Nunivak Island
muskox population. Alaska Coop. Wildl. Res. Unit. Univ. Alaska Fairbanks. 92pp.
. 1978. Musk-ox. Pages 135-149 in J. L. Schmidt and D. L. Gilbert eds., Big game
of North America. Wildl. Manage. Inst., Washington, D. C. 494pp.
, and D. Knudson. 1971. Muskox and snow cover on Nunivak Island, Alaska.
Pages 50-62 in A. O. Haugen, ed. Snow and ice in relation to wildlife and recreation
symposium. Iowa Coop. Wildl. Res. Unit, Iowa St. Univ., Ames.
Marcum, C. L., and D. O. Loftsgaarden. 1980. A nonmapping technique for studying
habitat preferences. J. Wildl. Manage. 44: 963-968.
Mech, L. D., R. E. McRoberts, R. O. Peterson, and R. E. Page. 1987. Relationship of
deer and moose populations to previous winter's snow. J. of Animal Ecology.
56:615-627.
Messier, F. 1991. The significance of limiting and regulating factors on the demography of
moose and white-tailed deer. J. of Animal Ecology. 60:377-393.
Miller, R. G. 1980. Simultaneous statistical inference. Second ed. Springer Series in
Statistics. Springer-Verlag. New York. 299pp.
Neter, J., W. Wasserman, and M. H. Kutner. 1985. Applied linear statistical models.
Second ed. R. D. Irwin, Inc. Homewood, Illinois. 1127pp.
N.O.A.A. 1988. Climatic atlas of the outer continental shelf waters and coastal regions of
Alaska:Volume HI Chukchi-Beaufort Sea. Natl. Oceanic and Atmos. Admin., Natl.
Climatic Center, Asheville, North Carolina. 77
. 1989. Local climatological data: annual summary with comparative data of Barter
Island, Alaska, 1988. Natl. Oceanic and Atmos. Admin., Natl. Climatic Center,
Asheville, North Carolina.
O'Brien, C. M. 1988. Characterization of muskox habitat in northeastern Alaska. M. S.
Thesis, Univ. Alaska Fairbanks. 114pp.
Owen-Smith, N and P. Novellie. 1982. What should a clever ungulate eat? Am. Naturalist
119:151-178.
Parker, G. R. 1978. The diets of muskoxen and Peary caribou on some islands in the
Canadian High Arctic. Can. Wildl. Ser. Occas. Pap. No. 35.
Rapota, V. V. 1984. Feeding ecology of the Taimyr muskoxen. Pages 75-80 in D. R.
Klein, R. G. White, and S. Keller, eds. Proc. First Int. Muskox Symp., Biol. Pap.
Univ. Alaska Spec. Rep. No. 4.
Rieger, S., D. Schoephorster, and C. Furbush. 1979. Exploratory soil survey of Alaska.
U.S.D.A. Soil Conservation Service.
Reynolds, P. E. 1990a. Winter distribution, movements, and habitat use of muskoxen on
potential petroleum lease areas of the Arctic National Wildlife Refuge. Pages 87-95 in
T. R. McCabe, ed. Terrestrial Research: 1002 Area - Arctic National Wildlife Refuge,
Annual Progress Rep., 1989. U.S. Fish Wildl. Serv., Anchorage, AK. 168pp.
Reynolds, P. E. 19906. Population dynamics of muskoxen on the Arctic coastal plain:
productivity and dispersal as a natural regulator of population size in the 1002 area of
Arctic National Wildlife Refuge. Pages 87-95 in T. R. McCabe, ed. Terrestrial
Research: 1002 Area - Arctic National Wildlife Refuge, Annual Progress Rep., 1989.
U.S. Fish Wildl. Serv., Anchorage, Alaska. 168pp. 78
Robus, M. A. 1981. Muskox habitat and use patterns in northeastern Alaska. M. S.
Thesis, Univ. Alaska Fairbanks. 116pp.
SAS Institute Inc. 1985. SAS Users Guide: Statistics. Fifth ed. Cary, North Carolina.
956pp.
Skogland, T. 1978. Characteristics of the snow cover and its relationship to wild mountain
reindeer {Rangifer tarandus tarandus L.) feeding strategies. A rct Alpine Res. 10:569
580.
Smith, T. E. 1984. Population status and management of muskoxen on Nunivak Island,
Alaska. Pages 52-56 in D. R. Klein, R. G. White, and S. Keller, eds. Proc. First Int.
Muskox Symp., Biol. Pap. Univ. Alaska Spec. Rep. No. 4.
Spencer, D. L. and C. J. Lensink. 1970. The muskox of Nunivak Island, Alaska. J. Wildl.
Manage. 34:1-15.
Tener, J. S. 1965. Muskoxen in Canada. Can. Wildl. Serv. Mono. Ser. No. 2. Queen's
Printer, Ottawa. 166pp.
Thing, H. 1977. Behavior, mechanics and energetics associated with winter cratering by
caribou in northwestern Alaska. Biol. Pap. Univ. Alaska, No. 18. 41pp.
Thing, H. 1984. Food and habitat selection by muskoxen in Jameson Land, northeast
Greenland: a preliminary report. Pages 69-74 in D. R. Klein, R. G. White, and S.
Keller, eds. Proc. First Int. Muskox Symp., Biol. Pap. Univ. Alaska Spec. Rep.
No. 4.
Thomas, D. C., and J. E. Edmonds. 1984. Competition between caribou and muskoxen,
Melville Island, NWT, Canada. Pages 93-100 in D. R. Klein, R. G. White, and S.
Keller, eds. Proc. First Int. Muskox Symp., Biol. Pap. Univ. Alaska Spec. Rep. No.
4. 79
Todd, J. W., and R. M. Hansen. 1973. Plant fragments in the feces of bighorns as
indicators of food habits. J. Wildl. Manage. 37:363-366.
Tyler, N. C., and A. S. Blix. 1990. Survival strategies in arctic ungulates. Rangifer.
Special Issue No. 3:211-230.
Walker, D. A., W. Avecedo, K. R. Everett, L. Gaydos, J. Brown, and P. J. Webber.
1982. Landsat-assisted environmental mapping in the Arctic National Wildlife Refuge,
Alaska. CRREL Rep. 82-37, U.S. Army Corps of Engineers, Cold Regions Res. and
Eng. Lab., Hanover, N.H. 59pp.
. 1983. A hierarchical tundra vegetation classification especially designed for
mapping in northern Alaska. Pages 1332-1337 in Proceedings of the Fourth
International Conference on Permafrost. July, 1983, Univ. Alaska Fairbanks.
Washington, D.C.: National Academy Press.
White, R. G., F. L. Bunnell, E. Gaare, T. Skogland, and B. Hubert. 1981. Ungulates on
arctic ranges. Pages 397-483 in L. C. Bliss, J. B. Cragg, D. W. Heal, and J. J.
Moore, eds. Tundra ecosystems: a comparative analysis. Cambridge Univ. Press,
London.
Whitten, K. R. and R. D. Cameron. 1980. Nutrient dynamics of caribou forage on
Alaska's arctic slope, in E. Reimers, E. Gaare, and S. Skjenneberg eds. Proc. Second
Int. Reindeer/ Caribou Symp., Rpros, Norway, 1979.
Zar, J. H. 1984. Biostatistical analysis. Second ed. Prentice-Hall Inc., Englewood Cliffs,
N.J. 718pp. PERSONAL COMMUNICATIONS
Mark Biddlecomb. Graduate Student Alaska Cooperative Fish and Wildlife Reseach Unit
Univ. Alaska Fairbanks.
Martin Raillard. Graduate Student Dept of Botany. Erindale College. Univ. Toronto.
Ontario, Canada.
Patricia Reynolds. Wildlife Biologist. U.S. Fish and Wildlife Service. Arctic National
Wildlife Refuge. Fairbanks, Alaska.
Don Young. Wildlife Biologist. U.S. Fish and Wildlife Service. Alaska Fish and Wildlife
Research Center. Fairbanks, Alaska.
80 Table 10. Cross reference of vegetation/land cover types of four habitat studies on the coastal plain of the Arctic National Wildlife Refuge, Alaska.
Jingfors (1980) This Sludv Christiansen (1990) O'Brien (1988) and Robus(1981) Water Water No comparable category No comparable category Wet Sedge Aquatic Graminoid tundra Wet Sedge Tundra Wet Sedge Meadow Wet sedge tundra Wet Sedge Tundra Wet Sedge Meadow Moist/wet Sedge Complex Wet/moist Sedge Tundra Complex Wet Sedge Meadow Moist Dwarf Shrub/Wet Sedge Tundra Complex No comparable category Moist Sedge Moist sedge shrub tundra Moist Dwarf Shrub, Sedge Tundra Heath Polygon Tundra Hummocky sedge tundra No comparable category No comparable category Tussock Tundra Water track complex No comparable category No comparable category Tussock Tundra Moist Sedge Tussock, Dwarf Shrub Tundra Tussock Meadow A APPENDIX Shrub tussock tundra No comparable category No comparable category Shrub moss tundra No comparable category No comparable category ooShrub Tundra Shrub tundra Moist Forb, Shrub Bluff Tundra No comparable category High center polygon complex No comparable category No comparable category Riparian Shrub Riparian shrub tundra Dry/Moist Riparian Tundra Complex Riparian Terrace/Willow Bar and Creek Willow Thicket Dryas Terrace Dryas terrace Dry, Prostrate Shrub, Forb Tundra Riparian Terrace Dryas Ridge Dryas ridge Dry Partially Vegetated Ridgetop Barren TundraDry Ridge Partially Vegetated Partially vegetated tundra Dry Partially Vegetated Gravel Bars Riparian Gravel Bars and Riparian Terrace/Willow Bars Barren Ground Barren Ground Dry Gravel Bar Barren No comparable category Barren No comparable category Riparian Grass Forb No comparable category Moist Streamside Forb, Graminoid Tundra No comparable category Gravel Bar APPENDIX B
Table 11. Snow depth (cm) of transect in moist sedge tundra on Barter Island during March and April during the first year of study and four previous years and the deviatior from the five year mean.
March April # of S.D. # of S.D. Year Snow Depth from Mean Snow Depth from Mean 1985 17 -.71 21 -.65 1986 17 -.71 20 -.76 1987 19 -.47 29 .19 1988 25 .24 23 -.44 1989 37 1.65 43 1.66
Five Year Mean 23 S.D. 8.48 Mean 27.2 S.D. 9.50
82 Table 12. Vegetation types of nested halves of feeding zones used and their availability in the study area. All types with two or fewer observations were pooled. Each nested half is displayed separately before and after pooling.
Used (Half If Available______Pooled Data Veuetation Tvne Observed Expected Observed Expected Total A Water 0 1.7 21 19.3 21 Used (Half h Available B Wet Sedge 2 6.6 80 75.4 82 Tvne Observed F.xnected Observed Expected Total C Moist Sedge 16 16.2 185 184.8 201 L Other 7a 15.0 180 172.0 187 D Tussock Tundra 16 9.8 106 112.2 122C Moist Sedge 16 16 2 185 184.8 201 T Shrub Tundra 1 3.3 40 37.7 41 D Tussock Tundra 16 9.8 106 112 2 122 F Riparian Shrub 0 .9 11 10.1 11 G Drvas Terrace 4 2 0 21 23 0 25 Cl Dryas Terrace 4 .3 0 3.7 4 Totals 43 43 0 492 492.0 535 II Dryas Ridge 0 1.7 21 19.3 21 d.f.=3 I Partially Vegeta 1 .5 5 5.5 6 Chi-square= 11.1 J Barren Ground 0 .8 10 9.2 10 P<0.05 K R. Forb Gravel 3 1.3 13 14.7 16 a = Use < Availability (P loi simultaneous confidence intervals<: 005). Totals 43 43.0 492 492.0 53? appendix
oo Used (Half 2) Available Pooled Data u> Vecetatioii Tvne Observed F.xnected Observed Expected Total A Water 0 1.6879 2119.3121 21 Used (Half 21 Available Total B Wet Sedge 0 6.4299 80 73.5701 82 Community Tvne Observed F.xnected Observed Expected c C Moist Sedge 18 16.316 185 186.684 201 L Other 5b 14.9 180 170.1 185 D Tussock Tundra 16 9.8056 106 112.194 122 C Moist Sedge 18 16 3 185 186.7 203 E Shrub Tundra 1 3.2953 40 37.7047 41 D Tussock Tundra 16 9.8 106 112.2 122 F Riparian Shrub 0 0.8841 11 10.1159 11 G Drvas Terrace 4 2 0 21 23.0 25 G Dryas Terrace 4 0.3215 0 3.6785 4 Totals 43 43.0 492 492.0 535 H Dryas Ridge 0 1.6879 21 19.3121 21 d f.=3 1 Partially Vegeta 1 0.4822 5 5.51776 6 Chi-square=13.7 J Barren Ground 0 0.8037 10 9.19626 10 P<0.05 K R. Forb Gravel 3 1.286 13 14714 16 b = Use < Availability (P for siniulianeous confidence intervals < 0.05). Totals 43 43.0 492 492.0 535 84
Figure 21. Paired observations of vegetation type in each half of the feeding zone and nonadjacent zone (meso scale) for 11 categories. The chi-square test for symmetry (Agresti 1990) was calculated for a reduced matrix where any category with two or fewer observations in either a row or a column were pooled into an "other" category.
Nonadjacent Zone Half #1 Reduced Matrix Feeding Zone Half #1 A B C P E F fi H I J K A. Water Nonadjacent Zone B. Wet Sedge 1 1 Half #1 C. Moist Sedge 1 2 m 1 1 1 Feeding Zone Half #1L C D D. Tussock Tundra 1 7 2 1 3 ? L. Other m E. Shrub Tundra i C. Moist Sedge
Nonadjacent Zone Half #2 Reduced Matrix Feeding Zone Half #2 A B C D E F fi H I I K A. Water Nonadjacent Zone B. Wet Sedge Half #2 C. Moist Sedge 3 0 4 l 2 Feedine Zone Half #2 L C D G D. Tussock Tundra l l 9 l 2 L. Other 1 E. Shrub Tundra C. Moist Sedge 14 m F. Riparian Shrub D. Tussock Tundra 4 1 9 G. Dryas Terrace l * G. Dryas Terrace 1 m H. Dryas Ridge I. Partially Vegetated l d.f.=6 J. Barren Ground Chi-square=13.9 K. Riparian Grass Forb Z m P=0.03 Gravel Bar Total=40 42.5 % of the FZ-2 are of the same type as PNZ-2 85
Figure 22. Paired observations of vegetation type of feeding and adjacent zone halves (local scale). The chi-square test for symmetry (Agresti 1990) was calculated for a reduced matrix where any category with two or fewer observations in either a row or a column were pooled into an "other" category.
Adjacent Zone Half #1 Reduced Matrix Feeding Zone Half #1 A B C D E F G H I J K A. Water Adjacent Zone B. Wet Sedge l H alf#l C. Moist Sedge ? 3 9 l I Feeding Zone Half #1L C D G D. Tussock Tundra l 1i f l l \ j \ L. Other .f l _L E. Shrub Tundra I C. Moist Sedge 7 m F. Riparian Shrub D. Tussock Tundra 7 I# G. Dryas Terrace 1 2 m G. Dryas Terrace 2 ■mk H. Dryas Ridge I. Partially Vegetated i d.f.=6 J. Barren Ground Chi-square=12.5 K. Riparian Grass Forb z m P=0.05 Gravel Bar Total=43 48.9 % of FZ-1 are of the same type as PAZ-1
Adjacent Zone Half #2 Reduced Matrix Feeding Zone Half #2 A B C DFFG HI J K A. Water Adjacent Zone B. Wet Sedge Half #2 C. Moist Sedge 2 2 12 1 Feeding Zone Half #2 L C D D. Tussock Tundra 9 4 ? 1 L. Other m E. Shrub Tundra C. Moist Sedge 12 F. Riparian Shrub D. Tussock Tundra 7 m G. Dryas Terrace 4 H. Dryas Ridge I. Partially Vegetated 1 d.f.=3 J. Barren Ground Chi-square=13.0 K. Riparian Grass Forb 2 m P<0.05 Gravel Bar Total=42 52.4 % of FZ-2 are of the same type as PAZ-2 8 6
Figure 23. Paired observations of vegetation type selection in nested halves of feeding zones for 11 categories. The chi-square test for symmetry (Agresti 1990) was calculated for a reduced matrix where any category with two or fewer observations in either a row or a column were pooled into an "other" category.
Feeding Zone Halfalf Feeding Zone Half #2 Reduced Matrix #1 ABC PR F G H I J K A. Water Feeding Zone B. Wet Sedge 2 Feeding Zone Half Half #2 C. Moist Sedge 16 #1 L D Ct D. Tussock Tundra 16 L. Other •# E. Shrub Tundra C. Moist Sedge m F. Riparian Shrub D. Tussock Tundra f t G. Dry as Terrace 4 G. Dry as Terrace 4 H. Dry as Ridge I. Partially Vegetated d.f.=6 J. Barren Ground Chi-sauare=2.0 K. Riparian Grass Forb m P>0.919 Gravel Bar Total=43 95.4 % of FZ-1 are of the same type as FZ-2 APPENDIX D
Table 13. Coefficients of determination (r2) and significance levels (p) for regression analyses of the relationship between cover and biomass for major plant species on the coastal plain of the Arctic National Wildlife Refuge. From Felix et al. (1989).
Species N r2 P Carex aquatilis 18 0.39 <0.001 Eriophorum angustifolium 35 0.69 <0.001 E. vaginatum 34 0.65 <0.001 C. biglowii 19 0.30 0.015 Salix planifolia ssp. pulchra40 0.74 <0.001 Betula nana 18 0.13 0.144 Ledum palustre ssp. decumbens12 0.66 0.001 S. reticulata 6 0.92 0.002 Pedicularis spp. 43 0.01 0.628
87 APPENDIX E Table 14. Rank order of the most abundant species by total occurrences in muskox reeding, adjacent, and nonadjacent zones. Scale effects are based on randomized block ANOVA s on the cover of each species. If cover was greater in tne feeding zone, then difference is positive (+); if cover was greater in the paired adjacent or nonadjacent zone, then difference is negative (-). Total Adjacent Scale Local Scale Soecies < Occurrences Difference Effect Difference Effect 1 Litter Alone 128 -r P=0.0034 + P=0.0396 2 Litter 128 + P=0.0036 -T- P=0.0013 3 Dicranum sp. 112 4- P=0.1942 + P=0.1275 4 Eriophorum angustifolium 112 + P=0.9934 + P=0.4734 5 Dryas integnfolia 110 -r P=0.6472 + P=0.1895 6 Salix planifolia 101 - P=0.3715 “f P=0.7129 7 Salix reticulata 98 - P=0.2006 - P=0.4365 8 Hylocomnium splendens 91 - P=0.7518 + P=0.9046 Tomentypnumnitens 88 P=0.9249 - P=0.0814 10 Drepanociadus sp. ~H - P=0.4597 - P=0.8002 11 Salix phlebopnylla ^ 3 - P=0.4128 - P=0.0089 12 Carex aquatilis ' i - P=0.2717 P=0.406 13 Eriophorum vaginatum T0 + P=0.0191 T- P=0.2364 14 Peltigera apthosa 69 + P=0.2187 + P=0.4886 15 Carex bigelowii 67 4- P=0.4128 + P=0.104 16 Aulacomnium acuminatum 66 -f P=0.1376 + P=0.8465 17 Polygonum bistorta 66 - P=0.8129 - P=0.6463 18 Aulacomnium turgidum 65 - P=0.0006 + P=0.1089 19 Tussocks 64 + P=0.0167 + P=0.3666 20 Soil 59 - P=0.6425 - P=0.1771 21 crustose lichens 58 - P=0.3962 - P=0.4534 22 Polytricum junipennum 54 + P=0.292 - P=0.6822 23 Calliergon giganteum 52 + P=0.5254 + P=0.3677 24 Cetrana cucullata 50 + P=0.1946 + P=0.5866 25 Pyrola grandiflora 48 + P=0.1237 + P=0.9509 26 Thamnolia sp. 48 + P=0.8481 + P=0.9408 27 Betulanana 47 + P=0.2903 + P=0.9424 28 Cassiope tetragona 47 + P=0.4656 + P=0.7138 29 Vaccinium vuis-idaea 47 + P=0.72 - P=0.419 30 Bryum sp. 45 + P=0.4474 + P=0.4095 31 Pedasites frigidus 45 - P=0.4578 + P=0.2197 32 Ptilidium ciliare 44 + P=0.9886 - P=0.1995 33 Ledum palustre 41 + P=0.5142 + P=0.9283 34 Equisetum variegatum 40 - P=0.9729 - P=0.9196 35 Water 40 - P=0.3259 - P=0.8845 36 Vaccinium uliginosum 37 - P=0.2305 - P=0.2478 37 Stellariasp. 36 + P=0.954 + P=0.6862 38 Aulacomnium palustre 35 + P=0.5704 + P=0.5943 39 Hummocks 34 - P=0.598 - P=0.3835 40 Rubus chamaemorus 33 - P=0.6829 + P=0.4574 41 Sphagnum sp. 32 + P=0.4534 + P=0.4611 42 Distichum capillaceum 30 + P=0.7982 + P=0.2036 43 Equisetum arvense 30 - P=0.0411 + P=0.1063 44 Rock 29 - P=0.007 - P=0.0381 45 Salix alaxensis 29 - P=0.5269 - P=0.7 111 46 Saxifraga punctata 28 - P=0.8865 + P=0.9613 47 Gravel 27 - P=0.023 - P=0.005 48 Astrasalus umbellatus 25 + P=0.5384 - P=0,3069 88 Table 15. Composition of late winter muskox feces collected in feeding /.ones in the Arctic National Wildlife Refuge as determined by microhistological analysis of plant fragments.
Percent Diet Composition Year Location______Sedges Grasses Willows Shrubs Horsetails Forbs Mosses Lichens OthieF 1989 Angun Point 35.6 ~TT3. '8.3 "■ 1.9 .0 7.3 13.6 2.0 .0 1989 Kogopat River 50.4 4.3 6.3 4.4 .0 4.3 30.1 .3 .0 1989 Kongakut River 34.3 9.8 17.9 4.9 .0 4.5 27.3 1.4 .0 1989 Ekaluakat River 43.3 7.7 9.6 2.3 .0 8.9 26.1 2.3 .0 1990 Beaufort Lagoon 61.3 12.3 .2 6.3 .0 6.2 10.8 2.0 .9 1990 Wolf Creek 15.2 58.1 9.1 2.6 .7 5.6 6.4 .3 2.1 1990 Kongakut River 41.9 6.9 12.1 4.2 .8 3.6 29.1 .4 .9 1990 Upper Jago 24.5 4.4 12.5 6.2 1.6 15.9 30.5 4.4 .0 1990 Lower Jago 25.8 10.7 6.8 3.8 .0 4.4 47.6 .9 .0 Mean 39.1 13.9 9.2 4.1 .3 6.7 24.6 1.6 .4 S.E. 5.1 5.6 1.6 .5 .2 1.3 4.2 .4 .2 F APPENDIX
00v£>
Table 16. Composition of winter-type muskox feces collected in the Arctic National Wildlife Refuge from 1988 to 1990 as determined by microhistological analysis of plant fragments. These were samples resembling winter pellets collected in summer.
Percent Diet Composition Location______Sedges Grasses Willows Other Shrubs Horsetails Forbs Mosses Lichens Other Sadlerochit River ..20.63" 12.15 .. 16.38 5.74 2.74 17.49 21.47 3.4 0 Camden Bay 22.06 3.95 11.11 0.39 0 7.13 49.85 5.51 0 Nularvik Bluffs 18.5 9.87 16.75 3.44 0.1 0.02 40.4 6.92 0 Pokok Bay 32.46 7.72 12.46 3.51 0 8.11 26.57 9.17 0 Kongakut River 23.83 5.59 34.21 5.08 6.88 7.09 17.32 0 0
Mean 233 " 7.9 18.2 3.7 1.9 8.0 31.1 5.0 0.0 S.E. 2.4 1.5 4.2 0.8 1.3 2.8 6.1 1.6 0.0 Table 17. Percent diet composition of muskoxen killed in late winter by hunters on the coastal plain of the Arctic National Wildlife Refuge as determined by microhistological analysis of rumen contents.
Date Percent Diet Composition Sex Killed Location Willows Shrubs Forbs Sedges Grasses Horsetails Mosses Lichens Other Bull Mar-89 Sadlerochit 3.98 11.46 14.24 25.52 22.6 0 21.47 0.73 0 Bull Mar-89 Sadlerochit 23.07 3.67 15.04 22.65 19.57 0 13.93 2.07 0 Bull Mar-89 Sadlerochit 11.88 2.93 9.41 26.6 29.42 0 15.8 3.96 0 Bull Mar-83 Canning- 5.26 0.22 36.98 17.4 23.05 0 13.8 0.31 2.98 Katakturuk Bull Mar-83 Canning- 0.21 0.13 12.25 40.85 28.77 0 16.69 1.1 0 Katakturuk Bull Mar-83 Canning- 3.61 0.17 10.28 17.57 4.15 58.29 4.58 1.35 0 Katakturuk Bull Mar-8 8 Sadlerochit 12.35 1.99 7.24 17.41 28.1 0 18.69 13.95 0 Mean 8.6 2.9 15.1 24.0 22.2 8.3 15.0 3.4 .4 S.E. 2.94 1.52 3.80 3.18 3.32 8.33 2.01 1.82 .43
Cow Apr-89 Niguanik 5.41 2.81 8.71 49.46 9.58 0 21.11 2.92 0 Cow Oct-82* Sadlerochit 8.27 13.03 2.38 61.56 3.76 0 8.37 1.49 1.14 Mean 6.8 7.9 5.5 55.5 6.7 .0 14.7 2.2 .6 S.E. 1.43 5.11 3.17 6.05 2.91 .(H) 6.37 .72 .57 *Died in poor condition
o